JP2004509611A - Visual servo controlled optical microscope - Google Patents

Abstract

The present invention provides a method and apparatus for discovering and optimizing differences between cell types in a knowledge base. In particular, the present invention provides a visual servo controlled optical microscope and analysis method. The present invention provides a means to closely monitor hundreds of individual living cells over time; quantification of dynamic physiological responses in multiple channels; real-time digital image segmentation and analysis; intelligent and repetitive, computer-applied cells. Stress and cell stimulation; and provide the ability for cells to return to the same field of view for long-term research and observation. The invention further provides a means of optimizing culture conditions for a particular subpopulation of cells.

Description

汎関数 Ｈ_Ｎ（ｆ） を最小にするオイラーの方程式は次式（式２）で表される。
【数２】

式中、下付き文字は例えば次式のような導関数を表す。
【数３】

第一の係数を消去し、かつＮ→∞とすることにより、式３が得られる。
【数４】

式（３）は無限ラプラス方程式（ＩＬＥ）と呼ばれるもので、広く研究されている（例えば、Ａｒｏｎｓｓｏｎ、Ａｒｉｋｖ． Ｍａｔｅｍａｔｉｋ、６：５５１−５６１［１９６６］； Ａｒｏｎｓｓｏｎ、Ａｒｉｋｖ． Ｍａｔｅｍａｔｉｋ、７：１３３−１５１ ［１９６７］； Ａｒｏｎｓｓｏｎ、Ｍａｎｕｓｃｒｉｐｔａ Ｍａｔｈ．、４１：１３３−１５１［１９８１］； ならびに、Ｅｖａｎｓ、Ｅｌｅｃｔｒｏｎ． Ｊ． Ｄｉｆｆｅｒｅｎｔ． Ｅｑｕａｔ．、［ｈｔｔｐ：／／ｅｊｄｅ．ｍａｔｈ．ｓｗｔ．ｅｄｕ／Ｖｏｌｕｍｅｓ／１９９３／０３−Ｅｖａｎｓ／ａｂｓｔｒ．ｈｔｍｌ］３：１−９ ［１９９３］を参照）。式（２）のいくつかの重要な特性は、少なくとも１つの解が存在すること、および、妥当に弱い意味で「解」が再定義されるならばＣ^１の連続解が存在することである。さらに、ｆの勾配の軌跡は凸曲線または直線であり、かつ、Ｒには停留点｜∇ｆ｜＝０が存在しない。局所座標系が∇ｆで定義される局所座標系において、式（３）がゼロ交差（ｆ_ｘｘ＋ｆ_ｙｙ＝０）と等価であることは興味深い。
【０１０４】
式（３）を生データに適用すると、ラプラス形式より安定な（すなわち、前景と背景との間の漏れがない）ゼロ交差マップが得られる。このゼロ交差マップは、強度制限条件およびサイズ制限条件で容易にフィルタリング可能な誤領域を有する閉鎖輪郭をもつ。フィルタリングされたセグメント画像は次に重心変換に使用される。
【０１０５】
ｂ．テンソルベースの特徴解析
ベクトル場解析は、数学および物理学などの古典的分野とともにコンピュータの視覚およびパターン認識においてもよく研究されている技術である。運動ベースの用途向けに多数の概念およびツールが開発されている。本発明は、現在の最新技術を科学画像のセグメント化および解釈に拡張する。強度画像ｆ_０（ｘ，ｙ）が与えられたとき、強度画像から導出される自然ベクトル場が存在し、これはπ／２の回転をもつ傾斜場に相当する：
【数５】

ここでの問題点は、ベクトル場「ｖ」にノイズが多く何らかの種類の正規化を導入する必要があることである。これは次式で表される：
【数６】

式中、ｕは正規化されたベクトル場である。次に、反復法により楕円ＰＤＥの解が得られる。次式の単純な線形（ガウス）目盛空間モデルにより一次近似を求めることができる（Ｌｉｎｄｅｂｅｒｇ、Ｊ．Ａｐｐｌ． Ｓｔａｔ、ｐａｇｅｓ ２２５−２７０ ［１９９４］）：
【数７】

式中、
【数８】

は標準偏差が
【数９】

であるガウス核である。
【０１０６】
次に、ベクトル場の特異点は、密なベクトル表現のコンパクトな抽象を与える。これらの特異点は点特徴または線形特徴により特徴付けることができる。点特徴には速度および鞍点が含まれる。これらのパターンが検出された後は、対応するオブジェクトの抽出が可能となる。ＲａｏおよびＪａｉｎ（ＩＥＥＥ
Ｔｒａｎｓ． Ｐａｔｔ． Ａｎａｌ． Ｍａｃｈ． Ｉｎｔｅｌｌ．、ｐａｇｅｓ ２２５−２７０ ［１９９４］）は、基礎となるベクトル場からの特徴抽出に局所ヤコビアンを用いることによって特異点を検出した。しかし、本発明のアプローチはより強力であり、かつ、渦のサイズ計測も可能である。渦サイズは、凸ブロブが互いに接続または接触している場合であっても凸ブロブの位置特定を補完する。これは、基礎となるベクトル場における特異点の位置を特定するジョルダンインデックス（Ｉｎｄｅｘ）に基づく。Ｆ＝（ｕ，ｖ）をベクトル場とし、Ｊを臨界点をもたないジョルダン曲線とする。Ｊのインデックスは次式により定義される：
【数１０】

各点ＰにおいてＰ周囲の半径Ｒの小円（Ｊ^Ｒ _Ｐによって表される）が選択され、Ｉｎｄｅｘ（Ｊ^Ｒ _Ｐ）が計算される。次に、以下のように流れ場（ｕ，ｖ）を分類することができる：
１．渦のインデックスが＋１ に等しい（すなわち、ベクトル場中の特異点は前述のＲａｏおよびＪａｉｎによって与えられる）場合、および
２．鞍点のインデックスが −１ に等しい場合。
次式のように、ベクトル場のダイバージェンスは全点においてゼロであるから、ベクトル場に節は存在しない：
【数１１】

次に、単純な検索技術により渦のサイズが推定される。点ａ（ｘ，ｙ）が渦であるならば、そのサイズＲ＊（ｘ，ｙ）は次式で定義される：
【数１２】

換言すると、Ｒ＊（ｘ，ｙ）は、Ｊ^Ｒ _{（ｘ， ｙ）} のインデックスが１であり続けるような最大のＲである。一次偏導関数の積分であるジョルダンインデックスは、ある意味において、強度画像の関数である。これとは対照的に、他の技術は、より多くのノイズが生じやすい高次の導関数を基礎としている。合成画像、蛍光染料で標識した核、および透過光で観察した細胞における渦および領域セグメント化の結果を種々の図に示した。
【０１０７】
図４１パネルＣおよびＤは、蛍光染料Ｈｏｅｃｈｓｔ ３３３４２で標識した、生細胞および溶解した細胞の核による結果を示す。図４１パネルＥおよびＦは、２０倍の位相差対物レンズと透過光とを用いて撮像した、生細胞による結果を示す。個々の細胞を位置特定する従来の技術は蛍光撮影に限られており、蛍光撮影は動的実験の設計の複雑さを増加させるため、後者の結果は重要である。本明細書に示すデジタル画像のうち、図４１パネルＥおよびＦのみは、１０×ＮＡ０．５ Ｚｅｉｓｓ Ｆｌｕａｒ対物レンズを用いずに撮影したものである。画像はすべて１０２４×１０２４であり、カメラ内のＣＣＤチップのピクセルは６．８×６．８μであった。したがって、１０倍で撮影したすべての画像において、視野はおよそ高さ７００μ、幅７００μである。
【０１０８】
ｃ．正規化した重心変換
Ｉ（ｘ，ｙ）を元の画像とする。各点（ｘ_０，ｙ_０）において、等高の輪郭は次式（式４）で定義される：
【数１３】

テイラー級数を用いて上式の拡張および打切りを行うと次式（式５）が得られる：
【数１４】

上式においてｕ＝ｘ−ｘ_０かつｖ＝ｙ−ｙ_０であり、または、
【数１５】

がヘッセの行列である次式の標準形（式６）
【数１６】

において
【数１７】

は強度の勾配であり、かつ、ｗ＝（ｕ， ｖ）^Ｔは局所座標系における重心である。式（６）で定義される二次曲線の重心が次式（式７）の線形制限条件を満たすことは周知である：
【数１８】

Ａが特異点でないならば、重心は直接求めることができる（すなわち、次式（式８））：
【数１９】

しかし、これは必ずしも真ではなく、一般的に次式（式９）で定義されるゼロ集合
【数２０】

は非自明である。さらに、これは以下の２つのカテゴリーに分類することができる。
１．強度の勾配がゼロである領域に対応する、均一な領域。二値画像の場合、情報は輪郭上のみに存在する；および、
２．グレーレベル画像中の楕円形状に対応する、不均一な領域。
【０１０９】
これらの点の重心は十分に定義されない。さらに、計算の安定性の観点からは、特異点のためこれらの近い点を高い信頼度で計算することができない。これらの難点に対処するためには問題を正規化する必要がある。（ｘ，ｙ）の重心を（ｕ（ｘ，ｙ），
ｖ（ｘ，ｙ））^Ｔで表すと、正規化モデルは以下の式１０または式１１で表すことができる：
【数２１】

または
【数２２】

式中、第１項および第２項は見積誤差、第３項は制限条件の平滑度であり、α（＞ ０）は重さである。この問題の解は「正規化重心変換」（ＲＣＴ）と呼ばれる。変分問題である式１１のオイラー−ラグランジュ方程式は次式（式１２）である：
【数２３】

【０１１０】
偏導関数の差分近似を上述の差分方程式に代入すると、次式（式１３）が得られる。
【数２４】

上式は次式（式１４）に書き直すことができる。
【数２５】

式中、（式１５）
【数２６】

である。
これらの係数はすべて偏導関数の関数であり、かつ、（ｕ（ｘ，ｙ）， ｖ（ｘ，ｙ））の近傍である。行列式（式１６）：
【数２７】

は常に正であり、かつ、式（１４）の解は次式（式１７）となる。
【数２８】

したがって、推定された偏導関数および前の推定値（ｕ^ｎ，ｖ^ｎ）から得られた推定値の新しい集合（ｕ^ｎ＋１，ｖ^ｎ＋１）は次式（式１８）となる。
【数２９】

【０１１１】
ｅ．表現および分類
生細胞の撮像、および一般的に蛍光顕微鏡撮像は、構造情報と機能情報とを分離するために多スペクトルであることが多い。いくつかの態様においては、試料を蛍光染料で標識し、３６０ｎｍでの撮像により核の情報（例えば、形状および構成）を得る。反応は他の励起波長（例えば、４９０ ｎｍおよび５７０ ｎｍ）で撮像される。本発明のいくつかの態様においては、それぞれの核は楕円および超二次関数で表現され、その反応は他のチャネルで直接読み取られる。
【０１１２】
楕円のフィッティングは、４ａｃ−ｂ^２＝１の制約条件を受ける多項式Ｆ（ａ，ｘ）＝ａｘ^２＋ｂｘｙ＋ｃｙ^２＋ｄｘ＋ｅｙ＋ｆのパラメータの推定に基づいて行われる（Ｆｉｔｚｇｉｂｂｏｎら、Ｐｒｏｃ．Ｉｎｔｌ． Ｃｏｎｆ． Ｐａｔｔ． Ｒｅｃｏｇｎ．、２５３−２５７ ［１９９６］）。
【０１１３】
２Ｄの超二次関数（Ｈａｎｓｏｎ、Ｃｏｍｐ． Ｖｉｓｉｏｎ Ｇｒａｐｈ． Ｉｍａｇｅ Ｐｒｏｃ．、４４：１９１−２１０ ［１９８８］； および、Ｋｕｍａｒら、ＩＥＥＥ
Ｔｒａｎｓ． Ｐａｔｔ． Ａｎａｌ． Ｍａｃｈ． Ｉｎｔｅｌｌ．、１７：１０７９−１０８３ ［１９９５］）は、次式（式１９）で定義される閉曲線である。
【数３０】

γ_ｉ ＞ ０であることから（Ｒｏｓｋｅｌｌｅｙら、Ｃｕｒｒ． Ｏｐｉｎ． Ｃｅｌｌ Ｂｉｏｌ．、７：７３６−７４７ ［１９９５］）、次式（式２０）が得られる。
【数３１】

上式は各ｉの平行線セグメントの対に対応する。これらの線セグメントは、その中に超二次関数が制限される凸ポリトープを（大きいγについて）定義する。この表現は、対称でなくてもよい幅広い形状について有効である。パラメータＡ_ｉおよびＢ_ｉは結合線の傾斜を決定し、これら２つの間の距離γ_ｉとＣ_ｉとは形状の「直角度」を決定する。
【０１１４】
フィッティング問題は以下のとおりである。ｎ個のセグメント
【数３２】

のｍ個のデータ点Ｐ_ｊ＝（ｘ_ｊ， ｙ_ｊ）、ｊ＝１， ２， ． ． ． ， ｍが与えられたとする。費用関数は次式（式２１）で定義される：
【数３３】

式中、
【数３４】

∇は勾配オペレータ、λは正規化パラメータ、Ｑ_ｉは制約条件項である（Ｋｕｍａｒら、ＩＥＥＥ Ｔｒａｎｓ． Ｐａｔｔ． Ａｎａｌ．Ｍａｃｈ． Ｉｎｔｅｌｌ．、１７：１０７９−１０８３ ［１９９５］）。パラメータＡ_ｉ、Ｂ_ｉ、Ｃ_ｉ、およびγ_ｉは、レベンバーグ−マーカー（Ｌｅｖｅｎｂｅｒｇ−Ｍａｒｑｕａｒ）の非線形最適化法（Ｐｒｅｓｓら、「Ｃにおける数値レシピ（Ｎｕｍｅｒｉｃａｌ
Ｒｅｃｉｐｅｓ ｉｎ Ｃ）」、Ｃａｍｂｒｉｄｇｅ Ｕｎｉｖｅｒｓｉｔｙ Ｐｒｅｓｓ ［１９９２］）を用いて、適切な初期推測値からεを最小化することによって計算される（Ｋｕｍａｒら、ＩＥＥＥ
Ｔｒａｎｓ． Ｐａｔｔ． Ａｎａｌ． Ｍａｃｈ． Ｉｎｔｅｌｌ．、１７：１０７９−１０８３ ［１９９５］）。画像中のそれぞれの核はルーメン中の位置によってさらに分類される。図８に、画像中の核の楕円フィッティングおよび分類の例を示す。
【０１１５】
ｆ．正規化した重心変換の解析
本発明の１つの態様において、画面内の核の抽出は周知の形態学的オペレータと異なる。１つの具体的な技術は、画像を三次元構造として捉える分水界変換である（例えば、ＮａｊｍａｎおよびＳｃｈｍｉｄｔ、ＩＥＥＥ
Ｔｒａｎｓ． Ｐａｔｔ． Ａｎａｌ． Ｍａｃｈ． Ｉｎｔｅｌｌ．、１８：１１６３−１１７３ ［１９９６］を参照）。画像をその局所極小からフラッドさせ、かつ異なる局所極小に由来する領域の併合を抑止することによって、画像は、流域（ｃａｔｃｈｍｅｎｔ）および盆地（ｂａｓｉｎ）、ならびに分水界線（ｗａｔｈｅｒｓｈｅｄ
ｌｉｎｅ）に分割される。別の態様では、このような分割は、各エッジ点から局所極小への下流経路を発見することによってエッジから開始される。分水界変換は、隣接領域を併合するために複雑な形態学的オペレーションを必要とする過剰セグメント化につながる。対照的に、本発明の好ましい態様では、二値化した画像から開始し、続いて、核に属する境界点をつぶして単一の点にする。ただし、自然の分割を明らかにするために境界点を局所盆地に移動させるという観点からは、分水界変換および重心変換も同じ発想である。この方法を用いると、距離変換に対応するエネルギーランドスケープは多数の局所極小を示す。対照的に、正規化重心変換は、単一の局所極小をもつ滑らかなエネルギーランドスケープを有する。
【０１１６】
重心変換は本質的に、二値ブロブを、各凸領域が核または他の副区画に対応するような別個の凸領域に分割する。分割は進化的であり、かつ、自然境界に沿って生じる。対照的に、本発明の開発過程で実施された早期の研究では（例えば、ＣｏｎｇおよびＰａｒｖｉｎ、Ｐａｔｔ．Ｒｅｃｏｇ．、３３：１３８３−１３９３ ［２０００］を参照）、このような分割は必ずしも対象の自然境界に沿って生じなかった。
【０１１７】
したがって、本発明は、生物学的反応（例えば生物学的機能）を惹起する種々の刺激（例えば外因的に印加された刺激）の結果として細胞間に発生するシグナルのより詳細な像を得るのに有用な、鏡検法および画像分析に対するバイオインフォマティクスアプローチのための方法を提供する。これらの方法は後により詳細に説明する。
【０１１８】
重要な点として、本発明のＶＳＯＭは、本発明者らの以前の研究でより詳細に報告されている分散型の鏡検法とインフォマティクスとのためのチャネルであるＤｅｅｐＶｉｅｗとのインターフェースに適している。Ｄｅｅｐ Ｖｉｅｗは、任意の機器をそのフレームワークに追加するための拡大縮小可能なインターフェースである。Ｄｅｅｐ
Ｖｉｅｗはまた、ワイドエリアネットワーク上でデータを転送および閲覧するための手段も提供する。システムは、その特性の広告（ａｄｖｅｒｔｉｓｅｍｅｎｔ）を介して標的の機器と相互作用する総称ＧＵＩを有する。ＶＳＯＭの場合、関係する特性はコンピュータ制御のシリンジ、シャッター、およびＸＹステージ、Ｚ軸フォーカスモーター、フィルタホイール、ならびにカメラ制御パラメータである。この、機器の広告性能（機能性）の概念は、より一様なインターフェースをもたらし、かつ、ソフトウェアが再使用可能であることにより維持管理費用を低減させる。
【０１１９】
Ｅ．セグメント化
前述のように、核および細胞質の自動描写は、細胞反応を特定の細胞区画にマッピングする上で重要な段階である。この計算構成要素はＶＳＯＭシステムと相互作用して１つまたは複数のチャネルから画像データを取得する（例えば、フィルタホイールを回転させるまたはシャッターを開閉させることによって）。単一のチャネルから取得した画像のセグメント化は前述のとおりである。以下に、多チャネルのセグメント化の概要を説明する。
【０１２０】
多チャネル画像の分析では、長時間を要する種類の、別のデータ収集が発生する。このシナリオでは、細胞の異なる構成要素を強調するため、異なる蛍光プローブを用いかつ異なる励起フィルタを使用して細胞が撮像される。この場合、システムは、複数のデータのチャネルの同時分析、特定の細胞内要素の選択、および繰り返し測定が可能でなければならない。ベイズフレームワークに基づいた融像のための効率的なアプローチが開発されている。本発明の開発中に行われた研究で示されているように（Ｐａｒｖｉｎら、ＡＡＡＩ
Ｓｙｍｐ． Ａｐｐｌｎ． Ｃｏｍｐ． Ｖｉｓ． Ｍｅｄ． Ｉｍａｇｉｎｇ ［１９９５］を参照）、このアプローチは（共焦点鏡検法を用いた）３Ｄにも拡張可能である。データ融合の目的は、セグメント化および標識のための異なる様式を補完的に処理することである。この見地から、セグメント化手順によりデータ領域中の各ピクセルが相応に標識される必要がある。しかし、標識処理を複雑化しうる曖昧さが多数存在する。これらの曖昧さは、局所処理のみから、および、高レベルのフィードバックの不在により、生じる可能性がある。
【０１２１】
曖昧さの発生源には、ノイズによるデータの破損、局所特徴を抽出するアルゴリズムの性能の限界、および、標識タスクを妨げる不必要な特徴の存在が含まれる。領域セグメント化の１つの局面では、ヒストグラムの分析により達成される、各領域の平均強度の推定が行われる。いくつかの態様において、ヒストグラムのピークの最初の位置が近似され、次に最小二乗近似により精密化される。計算過程の次の段階では、これらのピークをクラスター中心として使用して画像空間内の局所一貫性を高める。このとき、多チャネル情報に基づいて画像を標識するためにベイズフレームワークが使用される。具体的には、本発明は３つのレベルをもつベイズ階層モデルを使用する。第一レベルは基本的な分類Ｚのモデルである。理想的な条件下では、Ｍ個の異なる様式（例えば、青、緑、および赤の蛍光発光にそれぞれ対応する３つの蛍光画像）でＺを観察することにより、画像システム中の種々のノイズにより破損されるデータの理想的な表現Ｘ_ｉが特定される。階層の第三レベルはデータＹ_ｉの実際の観察結果に対応する。第一レベルは、マルコフランダム場（ＭＲＦ）を事前確率密度関数とともに、正のパラメータをもつイジング模型の形式で使用する。このパラメータは近傍ピクセル間の協力を促進する。ベイズフレームワークにおいて、画面の内容に関する事前の無知性（ｉｇｎｏｒａｎｃｅ）があるという仮定を設定してもよい。第二レベルは、Ｚが与えられたときの条件的な従属変数としてＸをモデル化する（すなわち、Ｐ（Ｘ_１，．．．，Ｘ_ＭＺ）＝Ｐ（Ｘ_１Ｚ） ．．． Ｐ（Ｘ_ＭＺ）。ある分類が細胞質としてラベル付けされるのであれば、緑チャネル（すなわち細胞質）および青チャネル（すなわち核）の理想的反応は互いに影響しないので、これは妥当である。階層の第三レベルであるＹ_ｉはガウス分布としてモデル化される。その結果、Ｙ_ｉはＸ_ｉの局所的にぼやけた表現となる。完全なモデルは次式（式２２）として表される：
【数３５】

最大事後確率（ＭＡＰ）推定を用いてＸｉおよびＺを得る。ただし、ＭＡＰ計算は本来的に実行不可能であるため、反復条件モード（ｉｔｅｒａｔｉｖｅ
ｃｏｎｄｉｔｉｏｎｉｎｇ ｍｏｄｅｓ； ＩＣＭ）に基づく数値近似を用いて局所最適条件を発見する（Ｂｅｓａｇ、Ｊ． Ｒｏｙ． Ｓｔａｔ． Ｓｏｃ． Ｂ、４８：２５９−３０２［１９８６］を参照）。
【０１２２】
Ｆ．ソフトウェア
本明細書に詳細に説明しているように、本明細書に記載の実験で使用したＶＳＯＭの機能アーキテクチャは、画像取得モジュール、画像分析モジュール、サーボ制御、およびアーカイブで構成される。画像取得モジュールは、６つの励起フィルタを用いて、微速度撮影の高分解能ビデオ鏡検法の手段を提供する。このモジュールは、種々の時間周波数および励起周波数の、手動の画像取得またはあらかじめプログラムされた画像取得を可能にするレシピマネージャを有する。レシピは、意味的な相互運用性を提供するＸＭＬ表記法で表現される。分析モジュールは、サイズ、位置、反応、および境界輪郭に基づいた画像の特徴ベース要約を提供する固有のセグメント化アルゴリズムを統合する。したがって、分析モジュールは、属性付けされたグラフモデルに基づいて画像の特徴ベース要約を提供する固有のセグメント化アルゴリズムのプールを有する。その結果、グラフ中の各節は画像中の同質の領域（例えば、核、細胞質など）に対応する。
【０１２３】
各節の属性には、境界輪郭、超二次関数を用いたこの輪郭のパラメトリック表現、派生する複数の特徴、および観察下における細胞内区画の反応が含まれる。グラフ中のリンクは種々の節の間の隣接関係をコードする。属性付けされたグラフモデルは、各細胞およびその近隣の構造的定義を完全に表現する。サーボ制御モジュールは、顕微鏡付近に位置する４つのシリンジのうち任意のシリンジの流れを制御することによって組織培養容器中の種々の化合物の濃度を制御する。サーボ制御は３つの操作モードを提供し、各モードはレシピマネージャで連続的に登録される。これらのモードには以下が含まれる：（１）各化合物の流量、その開始時点、および持続時間が特定される、静的レシピ；（２）観察下における細胞の特定の反応の関数として流量および持続時間を変化させる、動的レシピモード；ならびに、（３）指定された周波数で流れをオンおよびオフにするプログラムにより制御される、変調レシピ。アーカイブシステムは、画像、その計算されたグラフモデル、および解題データをフラットファイル環境に保存する。本発明の主要な特徴の１つは、複数の位置から遠隔的に操作できることである。この固有の特徴により、研究者らは地理的に離れた位置に物理的に位置していても、与えられた実験を協力して行うことができる。
【０１２４】
本発明により、ユーザーが利用できる柔軟性が最大限となる。例えばユーザーは、細胞チャンバへの灌流速度の制御、カメラ露出時間の設定、光学フィルタホイールの位置の選択、およびデータ収集のサンプリングレートの設定を可能にする、特定の実験のためのレシピを入力することができる。本明細書において、画像の取得および制御のための操作可能ＶＳＯＭソフトウェアパッケージの好ましい態様の１つを「ＶＸ５」と呼び、別の好ましい態様の１つを「ＡＤＲ」と呼ぶ。ＶＸ５はシステムの静的レシピ制御用に設計されたものであり、ＡＤＲは動的レシピの視覚サーボ制御実験を局所的にまたはネットワーク上で実行するために設計されたものである。図１４に、Ｄｅｅｐ
Ｖｉｅｗの機能的アーキテクチャの略図（パネルＡ）、ＶＸ５実行用のＸＭＬレシピファイルのサンプル（図１４パネルＣ）、ポンプログ中のデータの例（図１４パネルＢ）、および、ＩＣＳデジタル画像ヘッダの例（図１４パネルＤ）を示す。
【０１２５】
Ｇ．特徴ベースの画像の保存
現在、保存モデルは一般的にフラットファイル上に構築されており、フラットファイルは検索および維持が困難である。本発明は、保存取り扱いに必要なスキーマを構築するためのアクティブバイオインフォマティクスプログラムを活用する手段を提供する。１つのスキーマの設計を図１５に示す。この設計は、ＶＳＯＭに適応させるため、および、印加された刺激に対してどの特定の細胞内区画がどの程度「反応」したかというモデルを構築するうえで有用な、特徴ベースの空間−時間データベースをつくるために拡張されたものである。図１６に示す多次元データベースは、各細胞内区画の反応曲線を含む細胞反応と、１つの区画から次の区画への蛍光プローブの移動との表現を提供する。
【０１２６】
好ましい態様の使用中、各実験は、標本容器（例えばぺトリ皿）中の特定の物理的位置に対応する標的フレーム（すなわち画像）を含む。細胞内の構造はセグメント化され、属性グラフとして解題データとともに保存される。次に、コンピュータ制御下で生細胞に刺激を印加しながら経時的に収集した一連のデジタル画像として、標的フレームが観察される。各区画の反応が記録され、かつ、細胞内へまたは細胞内区画間を移動する蛍光プローブの動きが追跡される。
【０１２７】
保存されたデータの機能的ビューは階層的である。最下レベルには、生の画像および画像の対応する物理的解題が保存される。次のレベルには、属性付けされたグラフモデル（すなわち構造）、対応する反応（すなわち機能）、および環境条件の動的解題（例えば、刺激の種類、流量、濃度など）が保存される。次のレベルには、化合物が種々の細胞内区画の中で再配分する際の化合物の軌跡とともに、細胞内の各区画の時間的発現が反応カーブとして要約される。これらの反応は観察中の全細胞について類似していなくてもよい。しかし、これらの反応は、特定の種類の細胞群または類似の生理学的状態にある細胞群に各クラスターが対応するようなクラスターを生成することが予測される。これらのクラスターは次の相関研究のために同定されなければならない。各クラスターは時系列事象に対応し、時系列事象は細胞の特定の細胞内区画に対応する。各クラスターにおけるこれら反応の平均的な挙動は単値分解により表現される。正式には、観察される反応は次式（式２８）で定義される多値関数である：
【数３６】

式中、ｆは時点ｔにおいて特定の区画で観察される平均蛍光強度、［Ｄｙｅ］は染料の濃度、Ｃｈａｎｎｅｌは励起フィルタの位置、Ｔは温度であり、Ｆｌｏｗは注入される化合物の流れの変化に対応する。以後の知識発見用に観察結果をデータベースに保存するため、この経時変化する情報を取得するためのスキーマが開発される。各実験サイクルは観察した各細胞についてフィンガープリントを生成する。これらの個々の反応から、画像空間内の細胞内活動のコンパクトな表現が得られ、これは細胞ごと、細胞母集団ごと、細胞種ごとなどに可視化することができる、
【０１２８】
Ｈ．細胞反応からの学習
本発明の学習技術は細胞種間の細胞特異的な差異を最適化および発見するために使用される。この文脈において、計算された反応とその組織構造および軌跡とは、最適化された細胞種識別を検索するためのその後の段階的な検索を容易にするためのモデルの開発を補助する。いずれの学習アプローチにおいても、その基礎には、ポリシーと、報酬と、価値と、環境のモデルとがある。モデル再構成の目的の１つは、種々のポリシーおよび報酬関数を評価することである。モデルはある細胞種と別の細胞種とで異なることが予測される。これらのばらつきはそれ自体のメリットについて重要であり、かつ、定量化すれば貴重な妥当性確認として利用できる。データベース内容は、類似性測度の評価、精密化、および局在化の根拠を提供する。類似性は状態および動作に基づいてクラスターを分析することにより測定される。これらのクラスターは、同値類を確率するための、種々の反応に関する共起統計の計算を補助する。または、細胞反応およびその対応属性（動的解題）によりあるシーケンス中の表現が生成される。この表現は次に、そのシーケンス中の共起統計を推移確率とともに構築するのに利用される。共起統計は、代表的な特徴の集合の二値木分類の仮説設定を補助する。この方法は、ひとつには、相関するシーケンスの検出および分類に有用である。さらに、一旦モデルが確立されれば、異常値を発見する手段が共起統計により得られる。推移確率は環境の知識を提供し、環境の知識は次に強化学習に使用することができる。さらに、細胞反応を、経時変化する線形系としてモデル化する手段が提供される。線形系は、系の挙動を特徴付けるための単純かつ強力な方法である。この文脈において、可変メモリ（遅れ）を有するシステムの状態空間表現が構築され、かつ、関連する行列のパラメータが推定および検証される。
【０１２９】
Ｉ．強化学習
この相では、どのように蛍光プローブの取込みを変動させるか、および、系統的な方法で細胞内の取込みを最適化するかを学習するための開始点として既知の知識が使用される。これは通常「強化学習」と呼ばれ、かつ、いくつかの態様においてはマルコフ決定過程（ＭＤＰ）としてモデル化される。特定のＭＤＰは、その状態（すなわち観察された反応）と、蛍光プローブの細胞内濃度の変更（例えば、生物学的阻害剤を注入する、生理学的に重要なイオンの培地内の濃度を変更する、など）を含む動作との集合によって定義される。このような系における報酬は、計算された反応の変化またはそのエピソード的勾配の変化によって測定可能である。さらに、いくつかの態様においては、次に、データベース中の同じアッセイで得られた観察結果をもとに（ＭＤＰの）推移確率が計算される。これらの推移確率は最適ポリシー（すなわちスケジュール）を可能にする技術を提供する。このようなポリシーは一般的に、動的プログラミング、モンテカルロ法、および時間差分学習（ＴＤ学習）として表現可能である。動的プログラミングは細胞反応のモデルを必要とする。このようなモデルはデータベースから仮説設定される。さらに、データベースが拡張するにつれてモデルへのアクセシビリティが向上する。モンテカルロ法はモデルを必要としないが、ステップ−バイ−ステップの増分計算に適していない。一方、時間差分学習はモデルを必要とせず、かつ、完全に増分的である。しかし、これら３つの方法はすべて本発明の態様に適用可能である。例えば、サンプルの推移確率（完全ではないもの）が存在するならば、１つの増分ステップ（すなわち１つのエピソード）においてモンテカルロ法が使用可能である。また、時間差分法はモンテカルロ法と動的プログラミングとの組合せである（すなわち、モデルなしで生の経験から直接学習し、次に、学習した推定に一部基づいて推定を更新する。）
【０１３０】
Ｊ．悪性および非悪性の乳房上皮細胞を識別する生理学的特徴を迅速に発見する方法
原発乳房腫瘍細胞のインビトロ増殖における近年の進歩により、細針吸引（ＦＮＡ）で採取した比較的少数の腫瘍細胞を培養により継代および増殖することが可能になっている（Ｌｉら、Ｃａｎｃ．Ｒｅｓ．、５８：５２７１−５２７４ ［１９９８］）。事実、悪性ヒト乳房上皮細胞（ＢＥＣ）の初代培養は大きく進歩している（Ｂａｎｄら、Ｃａｎｃ． Ｒｅｓ．、５０：７３５１−７３５７［１９９０］； Ｅｔｈｉｅｒら、Ｃａｎｃ． Ｒｅｓ．、５３：６２７−６３５ ［１９９３］； Ｄａｉｒｋｅｅら、Ｃａｎｃ． Ｒｅｓ．、５７：１５９０−１５９６［１９９５］； ＴｏｍｉｄａおよびＴｓｕｒｕｏ、Ａｎｔｉ−Ｃａｎｃ． Ｄｒｕｇ Ｄｅｓ．、１４：１６９−１７７ ［１９９９］； ならびに、ＢｒｏｗｎおよびＧｉａｃｃｉａ、Ｃａｎｃ．Ｒｅｓ．、５８：１４０８−１４１６ ［１９９８］）。
【０１３１】
悪性細胞に関連したインビボ環境は、少ない酸素（すなわち低酸素）、低いｐＨ、低レベルのグルコースレベル、および高レベルの代謝廃棄物の領域で構成されていると考えられている。培養中でこれらの条件をシミュレートすると、非悪性細胞はこれらの厳しいインビトロ環境条件で生存できないため、原発乳癌細胞の比較的純粋な母集団が分離できることが示されている（Ｄａｉｒｋｅｅら、前述）。悪性のＢＣＥは非悪性細胞より速く増殖すると広く考えられているがこれは誤解である。したがって、厳しい環境で悪性細胞の培養を提供することのもう１つの利点は、増殖速度が速い非悪性の上皮細胞が存在せずかつ腫瘍細胞を凌駕して増殖できないこことである。また、腫瘍細胞の薬剤耐性は厳しい微小環境のストレスに依存するものである可能性があり、かつ、厳しい微小環境のストレスの結果である可能性がある（ＴｏｍｉｄａおよびＴｓｕｒｕｏ、前述）。この理由から、薬剤耐性を正確に測定するには適切な微小環境で細胞に対するアッセイを行わなけらばならない可能性があると予測されていた。事実、最も成功したインビトロの化学感受性試験のいくつかでは、両端をシールし１４日間インキュベートした小さな毛管で培養した細胞が使用された。その結果得られた毛管内の微小環境は、サンドイッチ状態のカバーガラス法で一般的に得られるものよりさらに厳しい可能性が高い（前述のＤａｉｒｋｅｅらの文献を参照）。
【０１３２】
さらに、培養においては、得られる細胞（最大１０^７個）は元の腫瘍とよく類似しており、かつ、軟寒天での増殖も含め１つまたは複数の腫瘍表現型を示す。事実、乳癌中には最適以下の栄養枯渇環境が存在する（Ｄａｉｒｋｅｅら、前述、ＴｏｍｉｄａおよびＴｓｕｒｕｏ前述、ならびに、ＢｒｏｗｎおよびＧｉａｃｃｉａ、前述）本発明は原発乳癌標本の培養を向上させる手段を提供する。例えば、腫瘍細胞と正常細胞とは、インキュベータのＣＯ_２レベルの関数として、大きく異なるｐＨ挙動を呈すると予測される。また、各細胞腫が耐性を示すｐＨレンジも異なる可能性が高い。正味の効果としては、単一のインキュベータ内で（別々のフラスコ内で、各細胞腫が好む培地中で）これら異なる細胞種が同時に増殖できるような共通のＣＯ_２レベルは存在しないことになる。事実、正常細胞が耐性を示すｐＨレンジは癌細胞のそれよりはるかに狭いと予測される。低ｐＨ環境（＜７．１）を好む正常細胞株は見つかっていない。
【０１３３】
したがって、ヒト乳癌上皮細胞の増殖成功には ｐＨ ＜ ７．０が必要であるということが正しいならば、これまでのインビトロ化学感受性アッセイが成功しなかった理由も明らかである。ほとんどの培地および細胞インキュベーション条件はｐＨ＞ ７．０となるよう設計されてきた。したがって、ヒト腫瘍生検標本をこれらの条件下で培養すれば、付着および増殖するのは癌細胞ではなく正常細胞である。結果として、それとは知らずに癌細胞ではなく正常細胞に対して化学感受性試験を行うことになる。これまで、臨床試験でこれらのアッセイが失敗したのはこれが理由である可能性がある。ただし、本発明を使用するには機序の理解は必要ではなく、かつ、本発明はいかなる特定の機序にも限定されない。しかし、本発明は、正常細胞および異常細胞の形態学的形質転換をｐＨおよびその他の微小環境の局面の関数として（例えば、透過光中の細胞をセグメント化する能力を有するＶＳＯＭシステムによって）モニター、分析、および定量化する手段を提供する。
【０１３４】
Ｋ．細胞の特定視野の繰り返し観察および分析
本発明はさらに、マルチウェルプレートについて本明細書に説明したのと同様の方式で、閉鎖した組織培養フラスコ内の細胞の多数の視野を繰り返し観察および撮像する手段を提供する。透過光中で細胞をセグメント化できる能力を有することは、フラスコが顕微鏡ステージに戻されたときにＶＳＯＭシステムが細胞の同じ視野に戻り、新しい画像を取得し、かつ形態学的変化を定量化するよう、細胞の形状、位置、および相対的位置を保存できることを意味する。これは、細胞の数、相対位置、および形状に変化があった場合でもシステムが細胞の同じ視野を再度位置特定するような、視覚的サーボ制御のより機械的な用途の例である。
【０１３５】
したがって、本発明のいくつかの態様において、システムは、ガラス切り用の微細仕上げホイールを使用して滅菌培養皿に軽くエッチングされた、極めて精密なグリッド（「＋」）に対する相対的な位置の手段を提供する。培養皿は次に除去し、インキュベータに戻すかまたは適切な条件下で保管してもよい。細胞を再撮像するときには、培養皿上の「＋」を基準にしてシステムが較正され、座標（０，０）が「＋」の中心に設定される。ユーザーは、最初の位置合わせ用に、透過光下でエッチングを容易に見ることができる。図５２の略図に、培養皿内の細胞位置の基準となるフレームを確立する目的で、エッチングされた培養皿を使用する様子を詳しく示した。このように、各細胞の反応を、培養皿中のその細胞の物理的位置において解題するという目的が達成される。さらに、刺激などに対する細胞の反応を経時的にモニターするために細胞の同一視野を数時間後または数日後に観察することもエッチングにより容易になる。
【０１３６】
これらの実験で使用されるシステムは、ガラス切り用の微細仕上げホイールを使用して滅菌培養皿に軽くエッチングされた、極めて精密なグリッド（「＋」）に対して相対的にその位置を参照する。培養皿は次に除去し、インキュベータに戻すかまたは適切な条件下で保管してもよい。細胞を再撮像するときには、培養皿上の「＋」を基準にしてシステムが較正され、座標（０，０）が「＋」の中心に設定される。ユーザーは、最初の位置合わせ用に、透過光下でエッチングを容易に見ることができる。図５２の略図に、培養皿内の細胞位置の基準となるフレームを確立する目的で、エッチングされた培養皿を使用する様子を詳しく示した。このように、各細胞の反応を、培養皿中のその細胞の物理的位置において解題するという目的が達成される。さらに、刺激などに対する細胞の反応を経時的にモニターするために細胞の同一視野を数時間後または数日後に観察することもエッチングにより容易になる。
【０１３７】
したがって本発明は、他の細胞（すなわち正常な乳腺上皮細胞）を増殖上不利にしながら特定の細胞（すなわち原発乳癌細胞）を増殖上有利にするために細胞種間の差異を同定および活用する方法およびシステムを提供するだけでなく、生細胞の増殖率を細胞ごとに定量できるＶＳＯＭ蛍光アッセイも提供する。
【０１３８】
Ｌ．インビトロの薬剤試験および化学感受性試験の向上
本発明は、悪性細胞の増殖および分析する手段を向上させることに加えて、癌患者（例えば乳癌患者）の臨床成績を向上させるための患者志向型の治療法を容易にする、インビトロの薬剤試験および化学感受性アッセイも大幅に向上させる。
【０１３９】
次の名称集を用いて臨床的相関の要約を作成することができる：ＴＰ（真陽性、細胞がインビトロで感受性を示しかつ化学療法に反応する患者）、ＴＮ（真陰性、細胞がインビトロで耐性を示しかつ化学療法に反応しない患者）、ＦＰ（偽陽性、細胞がインビトロでは耐性を示すが臨床的には耐性を示す患者）、ＦＮ（偽陰性、細胞がインビトロでは耐性を示すが臨床的には反応を示す患者）、ＰＰＡ（陽性予測精度）＝ＴＰ／（ＴＰ−ＦＰ）、試験システムで感受性を示しかつ治療に反応した患者のパーセンテージ、および、ＮＰＡ（陰性予測精度）＝ＴＮ／（ＴＮ＋ＦＮ）、試験システムで感受性を示しかつ治療に反応しなかった患者のパーセンテージ。
【０１４０】
インビトロの試験結果と患者の反応とのこれまでの相関データでは、総合的なＰＰＡは７２％、ＮＰＡは９０％となっている。これは、感度（すなわち、臨床的に反応性の患者を検出する能力）８５％、および特異度（すなわち、非反応性の患者を検出する能力）８９％に相当する。これらの数値は４２６３名の患者を対象とした研究における７種類のアッセイの平均であり、結果は複数の個別の研究からプールされたものである（ＤｅＶｉｔａ、「癌：腫瘍学の原理と実践（Ｃａｎｃｅｒ：Ｐｒｉｎｃｉｐｌｅｓ ａｎｄ Ｐｒａｃｔｉｃｅ ｏｆ Ｏｎｃｏｌｏｇｙ）」、Ｌｉｐｐｉｎｃｏｔｔ−Ｒａｖｅｎ、Ｐｈｉｌａｄｅｌｐｈｉａ ［１９９７］）。これらのアッセイが患者の反応よりも患者の薬剤耐性を予測するのに優れていることから、これらの著者らは、これらのアッセイを「化学療法」アッセイではなく「薬剤反応性」アッセイと呼ぶことを提唱している。
【０１４１】
ロボット視覚技術である「視覚サーボ制御（ＶＳ）」として結合された、デジタル撮像蛍光鏡検法、デジタル画像のリアルタイム分析、およびバイオインフォマティクスデータベースは、デジタル画像の内容の分析に基づいた実験パラメータの動的操作を意味する。さらに、ＶＳＯＭは複数の視野を進んでおよび戻って走査する能力を有し、かつ、これにより生細胞の非常に多数の母集団をモニターする能力を有する。
【０１４２】
前述のとおり、ＶＳＯＭは、細胞単位で変化をモニターするために、関心対象の細胞の同一視野に繰り返し戻る能力をさらに提供する。この能力により、細胞で観察される早期の生理学的反応を、増殖かアポトーシスかまたはある細胞周期段階での停止かという細胞の将来の生物学的状態または最終運命と相関させることが容易になる。事実、本発明の１つの態様は最も望ましい予測アッセイを提供する。重要な点として、これらのアッセイは動的かつ適応的である。
【０１４３】
１つの態様において、適応的ＶＳＯＭ蛍光アッセイでは、インテリジェント的に選択されたコンピュータ制御の刺激に対する細胞の早期の生理学的反応が観察される。いくつかの好ましい態様において、これらの刺激および摂動は可逆的であり、かつ、細胞を特定の生物学的経路に永続的に付すことはない。例えば、細胞外の培地のｐＨを低下させながら、観察されるストレス反応に基づいて細胞の特定の部分母集団を同定する。システムのインテリジェント反応の１つとしては、関心対象の細胞が同定されるまで培地のｐＨを低下させ、次に、細胞に不可逆的な損傷が生じる前に細胞を高いｐＨに戻すことが考えられる。次に、関心対象の細胞が細胞単位で同定され、システムは、低ｐＨの環境に対するこの差分反応を最適化および／または利用するための追加の試験を実施するために使用される。
【０１４４】
本発明はまた、抗癌剤または薬剤の組合せによって利用できる生理学的特徴を探索するため細胞を繰り返しプロービングできる能力を提供する。したがって、本発明のＶＳＯＭ蛍光アッセイは、好ましい態様において、細胞の複雑な混合物を用いて細胞を迅速に同定する手段であって、蛍光プローブが細胞に対して非毒性であり、印加された刺激または摂動に対する特徴的な生理学的反応に基づいて細胞が特定され、かつ、刺激または摂動が細胞に対して不可逆的影響をもたない手段を提供する。事実、特に好ましい態様において、本発明の予測アッセイは、インテリジェント的に選択されたコンピュータ制御の刺激に対する一連の早期の生理学的反応を観察することを含む動的かつ適応的なＶＳＯＭ蛍光アッセイである。このアプローチは、このような刺激および摂動が可逆的でありかつ細胞を特定の生物学的経路に永続的に付さない場合はさらに強力である。比較的無害の刺激および摂動を生細胞に繰り返し印加できる能力により、抗癌剤または抗癌剤の組合せによって利用できる生理学的特徴を探索するため細胞を繰り返しプロービングすることが可能となる。本発明のＶＳＯＭ蛍光アッセイは細胞の将来の挙動および／または最終運命を良好に予測できる。例えば本発明は、細胞が増殖できるよう十分に刺激されたか、または、アポトーシスが必然となる点までストレスを受けたかを決定する手段を提供する。この能力は、細胞で観察される早期の生理学的反応と、細胞の将来の生物学的状態または最終的な生物学的運命（すなわち、増殖、アポトーシス、特定の細胞周期段階での停止など）とを相関付けるのに役立つ。
【０１４５】
本発明はさらに、以前に観察された生理学的な細胞反応の記録を含むオブジェクト指向データベースを提供する。このデータベースは、ＶＳＯＭアッセイの知識ベース制御を実施するうえで非常に有用な資源を提供する。このデータベースに遠隔的に（すなわち、任意のＶＳ使用可能顕微鏡を介して）アクセスできることにより、これらの機器のインターネット上での制御が可能となり、かつ、これらの機器およびそれを操作するために必要なコンピュータの性能にかかる費用を大幅に削減しながらこれらの機器の機能性を大幅に高めることができる。
【０１４６】
少なくとも３つのプロスペクティブ臨床試験において、癌患者の化学療法に対する反応と生存率とを向上させるためにインビトロ化学感受性試験を使用する試みがなされている。しかし、これらの試験では、インビトロ化学感受性試験の結果に基づいて化学療法を受けた患者の生存率向上は認められなかった。しかし、それぞれの患者は他のどの患者とも少なくともわずかに異なる。例えば、１人の患者がある薬剤に対して有害反応を示し、別の患者が副作用をまったく経験しないということもあり得る。患者が薬剤に対して有害反応を示す場合または患者の腫瘍が薬剤に反応しない場合、医師は、効果のある薬剤を同定するため、薬剤および／または治療レジメンを変更する。乳癌の場合、医師は幅広い可能性の中から薬剤（または２つ以上の薬剤）を選択肢、かつ、個々の患者の腫瘍に対して最も有効でありかつ個々の患者における副作用が最小である組合せを見つけなければならない。多くの場合、医師は各患者について一連の薬剤を試さざるを得ない（すなわち、これらは最適の治療レジメンを決定するためのインビボ薬剤反応試験である）。インビトロの薬剤反応試験には、適切な治療法を見つけるための一連の衰弱を招く試行錯誤を患者に対して行うことなしにより幅広い薬剤をより短時間で試験できるという利点が期待できる。したがって、４０年以上にわたって、患者の体から腫瘍細胞を取り出しそれをインビトロで試験するという試みが行われてきた。
【０１４７】
成功しなかったこれらの方法で使用された２つのインビトロ化学感受性アッセイは、ＨＴＣＡ（ヒト腫瘍クローニングアッセイ）およびＤｉＳＣ（分染細胞毒性アッセイ）である。これらのアッセイで使用されるインビトロの細胞培養および細胞増殖の技術、ならびに測定される生物学的エンドポイントは著しく異なる。ＨＴＣＡでは、細切した腫瘍組織を非常に小さい毛管に入れ両端を密封し１４日間インキュベートした培養物が必要である。悪性細胞と非悪性細胞とを含むこの混合物（細胞数０．２〜１．０×１０^５個）を毛管内に密封する前に、ヒトの血漿で観察されるピーク濃度の１０分の１に相当する濃度に調整した標準的な抗癌剤に細胞を１時間曝露する。次に細胞を０．３％寒天に懸濁し、１００μｌの毛管に密封した状態で３７℃かつ７％ＣＯ_２の環境に１４日間おく。１４日後に細胞を毛管から抽出し、顕微鏡を用いて手作業でコロニー数を決定する。最初に必要な腫瘍細胞数が少なくてすみ、かつ、有意数の細胞を有するコロニーの発生が改善されたことから、この毛管クローニングシステムが使用されていた。
【０１４８】
ＤｉＳＣアッセイは、試験に利用できる細胞数を増幅するために小さいポリプロピレンチューブ内で液体培地を用いて腫瘍細胞を４〜６日間培養するという非クローニング的なアッセイである。薬剤への曝露時間は薬剤により異なり、１時間〜４日間である。一般的に、薬剤濃度は経験的に決定され、ＨＴＣＡで使用される濃度より高い。例えば、ＨＴＣＡでは０．０４μｇ／ｍｌのドキソルビシンで１時間細胞を処理するが、ＤｉＳＣでは１．２ μ／ｍｌで１時間である。ＤｉＳＣアッセイでは、ファストグリーンを含む懸濁液中で死細胞を染色することによって細胞膜の完全性を評価する。アセトアルデヒドで固定したカモ赤血球（ＤＲＢＣ）を内部標準として培養物に添加し、混合物全体を細胞遠心分離（ｃｙｔｏｃｅｎｔｒｉｆｕｇｅ）すると、生細胞は無色、死細胞およびＤＲＢＣは緑色に染色された状態で顕微鏡スライド上に収集される。次に、生細胞を染色するためＨＥ（ヘマトキシリン−エオシン）でスライドを対比染色する。次に、熟練技術者が顕微鏡的に細胞を腫瘍細胞または正常細胞として同定する。次に、ＤＲＢＣに対する生細胞の比が決定される。
【０１４９】
上述の試験で得られるデータについて共通の一群の問題点が指摘されている。両アッセイとも手作業で労働集約的であり、定量化を人の判断に依存しており、かつ、これらのインビトロシステムでは一部の患者の腫瘍細胞が良好に増殖しないためしばしば結果が得られないことがある。このような理由のため、これらのアッセイに対する臨床医の認知度および受容度は低い。したがって患者および標本の集積が困難となっている。しかし、患者の腫瘍細胞に効果を示さない薬剤の同定に優れた「第二世代」の試験がこれらの試験に置き換わりつつある。このようにいくつかの薬剤を除くことによって、医師は患者でインビボの薬剤反応試験を行う際にこれらの薬剤を考慮に入れる必要がなくなる。これらの試験は、労働集約的であるものの長年利用可能となっている技術を基礎にしている。しかし、これらの試験は労働集約的であり、かつ、人為的ミスが生じる可能性を有する。
【０１５０】
これまでのこうした研究が患者の反応および／または生存率に劇的な向上をもたらせなかったのは種々の要因の結果である可能性があり、こうした要因としては、微小環境（例えば、腫瘍細胞が好む微小環境がインビトロで適切にシミュレートされていなかった）、個々の細胞の生理学的反応がアッセイ中に詳しくモニターされなかった、ならびに／または、薬剤の濃度、組合せ、および曝露時間が制限されておりかつ科学的に最適化されていなかった、などがある。
【０１５１】
しかし、本発明（すなわち「第三世代」の試験の１つ）の開発中に、より進んだ細胞培養技術と、ロボット視覚およびデジタル撮像蛍光鏡検法の分野に由来する革新的な技術とが検討された。本発明の視覚サーボ制御光学顕微鏡（ＶＳＯＭ）は、悪性細胞と非悪性細胞（例えば乳房上皮細胞（ＢＥＣ））とを識別する重要な生理学的特徴の迅速な発見などに有用である。前述のとおり、ＶＳＯＭは、関心対象の細胞（例えば腫瘍細胞）の増殖に有利に働くインビトロの微小環境を迅速かつ自動的に生成すること、および、印加されたストレスに対する早期の生理学的反応に基づいて細胞の挙動を予測するインビトロの蛍光アッセイを生成することができる。したがって、本発明は前述の臨床試験に関連する問題点を克服する手段を提供し、このような問題点としては、インビトロで腫瘍細胞を増殖させるうえでの困難、アッセイの過程で間隔をおいて細胞を観察する能力に限界があること、１回のアッセイで観察できる生物学的エンドポイントの数に限界があること、および、オートメーションの欠如が含まれる。本発明はまた、原発乳癌細胞（例えばヒト細胞）のインビトロ増殖と化学感受性試験とに適した微小環境条件を迅速に発見および最適化する手段を提供する。さらに本発明は、この目的を腫瘍ごとに達成する手段を提供する。
【０１５２】
Ｍ．微小環境パラメータのモニタリング
本発明は、ＶＳＯＭ実験に有用である重要な微小環境パラメータのオンラインモニタリングに必要なバイオセンサーおよび関連電子機器を使用する手段を提供する。これらの実験で使用されるセンサーは灌流ラインへの直列接続に適したＺＡＢＳフロースルーセンサーであると予測される。したがって、細胞に到達する直前の培地について複数の重要な特性を連続的にモニターおよび記録することが可能である。本発明のいくつかの態様においては、大きな透過光集光装置、温度プローブ、および真空吸引装置が存在するため、実際の細胞灌流チャンバ内にこれらのセンサーのすべてを設置するための空間が十分にない。したがって、培地をマイクロインキュベーションチャンバ内に連続的に灌流させながら、培地のｐＨ、Ｌ−乳酸またはグルコース、細胞外カルシウム、およびｐＯ_２をモニターする。
【０１５３】
いくつかの態様において、ＶＳＯＭ法では、コンピュータ制御された４つのシリンジ灌流ポンプが使用される。さらに、コンピュータ制御のＸＹ走査ステージにしっかり取り付けられるようマイクロインキュベーションチャンバを改修した。真空駆動吸引装置が余分の液体を排除し、温度プローブが温度制御用のフィードバックを提供する。
【０１５４】
種々の態様において、電子機器パッケージが本発明とともに使用される。例えば、電子機器パッケージ「ＰＬＵＧＳＹＳ」がこれらの実験で使用される。ＰＬＵＧＳＹＳはモジュール設計であり、その基本システムケースは、ｐＨ、Ｌ−乳酸またはグルコース、細胞外カルシウム、およびｐＯ_２等の測定に必要な種々の増幅モジュールを保持するためのスロットを十分に有する。ＰＬＵＧＳＹＳケースはまた、使用されるデータ取得用ハードウェアとの互換性も有し、このようなハードウェアとしてはＰＣ用のＰＣＩ
Ａ／Ｄ変換基板が含まれる。この装置は、ＶＳＯＭシステム制御用の重要なリアルタイムデータを提供し、かつ、ＶＳＯＭ実験の実施および解釈に重要な微小環境パラメータと、ＶＳＯＭ実験中に使用される重要な微小環境パラメータのオンラインモニタリングに必要な関連電子機器との連続的なモニタリングおよび記録を可能にする。
【０１５５】
実験
以下の実施例は、本発明の特定の好ましい態様および局面を説明しかつさらに例証することを目的として提供されるものであり、したがって本発明の範囲を限定すると解釈されるべきではない。
【０１５６】
以下の実験の開示では次の略語を使用する：ＶＳ（視覚的サーボ制御）；ＶＳＯＭ（視覚的サーボ制御光学顕微鏡／鏡検法）；℃（度、摂氏）；ｒｐｍ（毎分回転数）；ＢＳＡ（ウシ血清アルブミン）；Ｈ_２Ｏ（水）；ａａ（アミノ酸）；ｂｐ（塩基対）；ｋｂ（キロ塩基対）；ｋＤ（キロダルトン）；ｇｍ（グラム）；μｇ（マイクログラム）；ｍｇ（ミリグラム）；ｎｇ（ナノグラム）；μｌ（マイクロリットル）；ｍｌ（ミリリットル）；ｍｍ（ミリメートル）；ｎｍ（ナノメートル）；μｍ（マイクロメートル）；Ｍ（モル）；ｍＭ（ミリモル）；μＭ（マイクロモル）；Ｕ（単位）；Ｖ（ボルト）；ＭＷ（分子量）；ｓおよびｓｅｃ（秒）；ｍｉｎ（ｓ）（分）；ｈｒ（ｓ）（時間）；ＭｇＣｌ_２（塩化マグネシウム）；ＮａＣｌ（塩化ナトリウム）；ＯＤ_２８０（２８０ｎｍでの光学濃度）；ＯＤ_６００（６００ ｎｍでの光学濃度）；［Ｃａ^＋２］_ｉｎ（細胞内カルシウム濃度）；［Ｃａ^＋２］_ｅｘ（細胞外カルシウム濃度）；ＤＯＸ（ドキソルビシン）；ＢＥＣ（乳房上皮細胞）；ＨＭＥＣ（ヒト乳腺上皮細胞）；ＭＣＦ−７（ヒト乳癌細胞株の１つ）；ＭＣＦ−７ＷＴＣ（ＭＣＦ−７ 野生型、Ｃｏｗａｎ）；ＨＴＣＡ（ヒト腫瘍クローニングアッセイ）；ＤｉＳＣ（分染細胞毒性）；ＣＡＭ（カルセイン−ＡＭ）；ＤＲＢＣ（カモ赤血球）；ＤＳ（薬剤感受性）；ＭＤＲ（多剤耐性）；ＥＨ（エチジウムホモ二量体）；ＭＥＢＭ（乳腺上皮細胞基本培地）；ＭＥＧＭ（乳腺上皮細胞増殖培地）；ＰＢＳ（リン酸緩衝生理食塩水）；Ｄ＿ＰＢＳ（ダルベッコのＰＢＳ）；ＦＮ（偽陰性）；ＴＮ（真陰性）；ＦＰ（偽陽性）；ＴＰ（真陽性）；ＮＰＡ（陰性予測精度）；ＰＰＡ（陽性予測精度）；ＰＭＡ（ホルボール１２−ミリステート１３−アセテート）；ＴＲＭＥ（テトラメチルローダミンエチルエステル）；ＴＧ（タプシガルギン）；ＦＵＲＡ−２−ＰＥ３（カルシウム感受性の蛍光染料の１つ）；Ｈ４２（Ｈｏｅｃｈｓｔ（ヘキスト）３３３４２）；ＬＴ−Ｒ（Ｌｙｓｏｔｒａｃｋｅｒ−Ｒｅｄ）；ＦＮＡ（細針吸引）；ＰＡＧＥ（ポリアクリルアミドゲル電気泳動）；ＰＢＳ（リン酸緩衝生理食塩水［１５０ ｍＭ ＮａＣｌ、１０ ｍＭ リン酸ナトリウム緩衝液、ｐＨ ７．２］）；ＰＣＲ（ポリメラーゼ連鎖反応）；ＰＥＧ（ポリエチレングリコール）；ＳＤＳ（硫酸ドデシルナトリウム）；Ｔｒｉｓ（トリス（ヒドロキシメチル）アミノメタン）；ＤＭＳＯ（ジメチルスルホキシド）；ｗ／ｖ（質量体積比）；ｖ／ｖ（体積体積比）；Ａｍｅｒｓｈａｍ（Ａｍｅｒｓｈａｍ
Ｌｉｆｅ Ｓｃｉｅｎｃｅ， Ｉｎｃ．、イリノイ州Ａｒｌｉｎｇｔｏｎ Ｈｅｉｇｈｔｓ）；ＩＣＮ（ＩＣＮ Ｐｈａｒｍａｃｅｕｔｉｃａｌｓ， Ｉｎｃ．、カリフォルニア州Ｃｏｓｔａ
Ｍｅｓａ）；ＡＴＣＣ（Ａｍｅｒｉｃａｎ Ｔｙｐｅ Ｃｕｌｔｕｒｅ Ｃｏｌｌｅｃｔｉｏｎ、メリーランド州Ｒｏｃｋｖｉｌｌｅ）；Ｂｅｃｔｏｎ Ｄｉｃｋｉｎｓｏｎ（Ｂｅｃｔｏｎ
Ｄｉｃｋｉｎｓｏｎ Ｌａｂｗａｒｅ、ニュージャージー州Ｌｉｎｃｏｌｎ Ｐａｒｋ）；ＢｉｏＲａｄ（ＢｉｏＲａｄ、カリフォルニア州Ｒｉｃｈｍｏｎｄ）；Ｃｌｏｎｅｔｉｃｓ（Ｃｌｏｎｅｔｉｃｓ、Ｓａｎ
Ｄｉｅｇｏ）；Ｃｌｏｎｔｅｃｈ（ＣＬＯＮＴＥＣＨ Ｌａｂｏｒａｔｏｒｉｅｓ、カリフォルニア州Ｐａｌｏ Ａｌｔｏ）；Ｍｏｌｅｃｕｌａｒ Ｐｒｏｂｅｓ（Ｍｏｌｅｃｕｌａｒ
Ｐｒｏｂｅｓ、オレゴン州Ｅｕｇｅｎｅ）；ＧＩＢＣＯ ＢＲＬまたはＧｉｂｃｏ ＢＲＬ（Ｌｉｆｅ Ｔｅｃｈｎｏｌｏｇｉｅｓ， Ｉｎｃ．、メリーランド州Ｇａｉｔｈｅｒｓｂｕｒｇ）；Ｉｎｖｉｔｒｏｇｅｎ（Ｉｎｖｉｔｒｏｇｅｎ
Ｃｏｒｐ．、カリフォルニア州Ｓａｎ Ｄｉｅｇｏ）；Ｎｅｗ Ｅｎｇｌａｎｄ Ｂｉｏｌａｂｓ（Ｎｅｗ Ｅｎｇｌａｎｄ Ｂｉｏｌａｂｓ， Ｉｎｃ．、マサチューセッツ州Ｂｅｖｅｒｌｙ）；Ｎｏｖａｇｅｎ（Ｎｏｖａｇｅｎ，Ｉｎｃ．、ウィスコンシン州Ｍａｄｉｓｏｎ）；Ｐｈａｒｍａｃｉａ（Ｐｈａｒｍａｃｉａ， Ｉｎｃ．、ニュージャージー州Ｐｉｓｃａｔａｗａｙ）；Ｓｉｇｍａ（Ｓｉｇｍａ
Ｃｈｅｍｉｃａｌ Ｃｏ．、ミズーリ州Ｓｔ． Ｌｏｕｉｓ）；Ｓｔｒａｔａｇｅｎｅ（Ｓｔｒａｔａｇｅｎｅ Ｃｌｏｎｉｎｇ Ｓｙｓｔｅｍｓ、カリフォルニア州Ｌａ
Ｊｏｌｌａ）；Ｃｒｅａｔｉｖｅ Ｓｃｉｅｎｔｉｆｉｃ Ｍｅｔｈｏｄｓ（Ｃｒｅａｔｉｖｅ Ｓｃｉｅｎｔｉｆｉｃ Ｍｅｔｈｏｄｓ， Ｉｎｃ．、ｗｗｗ．ｃｒｅ８ｉｖｅ−ｓｃｉ．ｃｏｍ）；および、Ｚｅｉｓｓ（Ｃａｒｌ
Ｚｅｉｓｓ， Ｉｎｃ．、ニューヨーク州Ｔｈｏｒｎｗｏｏｄ）。
【０１５７】
実施例１
同じ細胞セットに繰り返し戻るための撮像プロトコル
これらの実験では、培養中の同じ細胞セットに繰り返し戻るための方法を説明する。これらの方法を用いることにより、可動顕微鏡ステージ上に置かれた細胞培養皿中の生細胞の１つまたは複数のセットを観察することが可能となる。重要な点として、これにより、培養皿をステージから除去し、細胞を増殖させるためなどの目的でインキュベーに入れることが可能となる。次に培養皿を顕微鏡ステージに載せ、同じ細胞セットを自動的に観察に供することが可能である。これらの段階は所望のまたは必要な回数繰り返すことができる。
【０１５８】
これらの実験では、種々の化合物を培養皿に灌流させながら細胞セット（倍率１０Ｘで細胞５００〜１０００個）の反応をモニターすることができる。指定された数のチャネルからなる画像のスタックが、指定された長さの時間にわたって指定された間隔で１回以上取得される。編集可能な（テキストの）レシピファイルにより、使用される実験レシピが指定される。「スナップショット」用には１画像のみのスタックが取得され、「微速度撮影」実験用には画像の完全なスタックを規則的な間隔で取得することができる。連続微速度撮影実験では実験と実験との間にユーザーがステージまたは培養皿を動かさないことが好ましいため、本発明により、培養皿を動かす前に同じ細胞セットに対して連続的に複数の微速度撮影実験を行うことを容易にする。所望の分析を行った後、培養皿は顕微鏡ステージから取り除かれ、必要に応じて数時間または数日間インキュベータ内に置かれる。培養皿を顕微鏡ステージに戻したとき、細胞は生きていてもまたは死んでいても（例えば、固定または染色された細胞であっても）よい。無論、死細胞を使用する場合は微速度撮影実験は行われない。ただし、これらの実験においては、繰り返しモニターされる単一の細胞セットが周囲の多くの細胞セットを代表していることを確認する目的で、細胞周囲の視野も「スナップショット」画像スタックを取得することによって撮像される。
【０１５９】
これらの実験では、ファイル名の中の＿０、＿１、＿２、＿３、＿４、＿５、および＿Ｘによって「チャネル指示子」が指定される。これらの数字は光学フィルタホイールの特定の位置を示し、＿Ｘは透過光画像を示す。フィルタホイールの各位置の光学フィルタの識別は、波長に応じて各画像のヘッダーに記述される。使用する透過光の種類は指定されない。例えば「ｄｃ１２２９９９ｄｃ２．０００１．１＿０ｉｃｓ」とはフィルタホイールの位置「０」で画像が取得されたことを表す。画像ヘッダーに基づくと、この位置のフィルタは３６０ｎｍの蛍光励起フィルタである。チャネル指示子「ｄｃ１２２９９９ｄｃ２．０００１．１＿Ｘ．ｉｃｓ」とは透過光シャッターを開放した状態で画像が取得されたことを示す。この画像は位相差像、ＤＩＣ、明視野像などであってもよい。これらの実験で使用されるフィルタホイールの各位置の光学フィルタの最大波長の識別は、後に詳述する編集可能なｃｏｎｆｉｇテキストファイルに記述される。
【０１６０】
「細胞セット」とは、特定の倍率でデジタル画像中の特定の視野内に見えるすべての細胞を指す。数時間および数日間の過程を通じて、単独の生細胞が大きくなり分裂して２つになる、細胞が崩壊する、細胞が死んで浮流する、および個々の細胞が移動するなどの可能性がある。細胞セットの最初の位置は、培養皿（例えばぺトリ皿）表面に物理的にエッチングされた「＋」（十字線）記号である（０，０）点に対する相対的な位置として特定される（図５２参照）。
【０１６１】
「画像スタック」とは、単一の時点における指定された全チャネルの集合である。例えば、画像スタックの微速度撮影系列中の７番目のスタックは以下のようになりうる：

【０１６２】
「レシピファイル」とは、使用するプロトコルを指定するテキストファイルである。例えば、レシピファイルは以下のものを含む：
どの蛍光チャネルが完全画像スタックに含まれるか；
各チャネルの露出時間；
画像チャネルが取得される順序；
微速度撮影実験で取得される画像スタックの間隔；
取得するスタックの数；
フィルタホイールの各位置の励起フィルタの識別。
【０１６３】
方法の実施中、複数回の反復が可能である。例えば、「初回の反復」ではこれまで撮影されていない細胞の培養皿が使用される。細胞の入った培養皿にエッチングされた十字線を使用して、コンピュータ画面上で（０，０）マークが設定される。ユーザーは観察する細胞セットを探す。この段階において、種々のシャッターが開閉され、フィルタホイールが種々の位置に動かされ、かつ、プログラムは（０，０）マークに対してユーザーが培養皿をどこに移動したかを追跡し続けなければならない。次にユーザーは関心対象の細胞セットに焦点を合わせる。次に、レシピファイルの情報に基づいて実験が行われる。この初回の実験は、典型的に、生細胞の微速度撮影アッセイとして行われる。
【０１６４】
しかし、他の態様では、固定または染色された細胞の一連の「スナップショット」が使用される。次に、細胞またはステージを動かすことなく（焦点を再度合わせるのは許容される）レシピファイルが再編集され、１つまたは複数の微速度撮影実験が行われる。次に、細胞の入った培養皿が取り除かれ、必要時までインキュベータ（または他の保管場所）内に置かれる。「スナップショット」を使用するいくつかの態様では、他の細胞セットを観察するためユーザーが培養皿を動かし、追加のスナップショットが撮影される。
【０１６５】
「２回目の反復」では細胞セットが再度観察される。この細胞セットはすでに撮影され、かつ、その位置に関する情報が保存されている。まず、分析する細胞の入った培養皿が顕微鏡ステージに置かれる。十字線マークの最初の透過光画像が取得される。ユーザーは画面上の（０，０）マークを「クリック」し、どの細胞セットを観察するかをプログラムに指示する。プログラムは視野を（０，０）マークに移動させ、次に適切な細胞視野に移動させる。ユーザーは細胞に焦点を合わせ、編集されたレシピファイルの情報に基づいて実験を行う。この段階において、ユーザーは焦点を合わせるためにシャッターを開きかつフィルタホイールを動かす。細胞またはステージを動かすことなく（焦点を再度合わせるのは許容される）レシピファイルが再編集され、１つまたは複数の追加の微速度撮影実験が行われる。スナップショット実験では、所望に応じて視野が変更され、かつ、所望の数の追加のスナップショットが取得される。実験プロトコルの最後として、細胞の培養皿がステージから取り除かれかつ保管場所（例えばインキュベータ）に戻される。「ｎ回目の反復」について、生細胞のセットが繰り返し撮像されかつ細胞位置に関する情報が保存される。事象の順序は２回目の反復と同様である。
【０１６６】
実施例２
ＶＳＯＭの使用
本実施例では、本発明のＶＳＯＭの装置およびその他の局面について説明する。
【０１６７】
Ａ．ＶＳＯＭシステム光学プラットフォーム
一般的に、顕微鏡が倒立型（すなわち対物レンズが上向き）の形状であるほうが細胞チャンバおよび細胞容器の設計が容易であるため、大多数の態様において、ＶＳＯＭシステムは倒立型の蛍光顕微鏡を中心に構成される。しかし、正立顕微鏡（対物レンズが下向きの顕微鏡）で使用できる細胞チャンバも存在する。本発明のＶＳＯＭシステムはこのような光学プラットフォームとともに使用するのに適している。カメラおよび顕微鏡周辺装置を顕微鏡に取り付けても安定するだけの十分な重量をもった、生物学研究に適した等級の蛍光顕微鏡が好ましい。いくつかのＶＳＯＭ実験には標準的な１０倍対物レンズで十分である。研究用等級の顕微鏡、対物レンズ、および種々の周辺装置は、カール・ツアイス（Ｃａｒｌ
Ｚｅｉｓｓ， Ｉｎｃ．、ニューヨーク州Ｔｈｏｒｎｗｏｏｄ）およびニコンＵＳＡ（Ｎｉｋｏｎ ＵＳＡ、ニューヨーク州Ｍｅｌｖｉｌｌｅ）など種々の製造業者により製造されている。環境制御細胞チャンバは、ツアイス、ハーバード・アパラタス（Ｈａｒｖａｒｄ
Ａｐｐａｒａｔｕｓ、マサチューセッツ州Ｈｏｌｌｉｓｔｏｎ）およびバイオプテクス（Ｂｉｏｐｔｅｃｈｓ， Ｉｎｃ．（Ｂｉｏｌｏｇｉｃａｌ Ｏｐｔｉｃａｌ
Ｔｅｃｈｎｏｌｏｇｉｅｓ）ペンシルベニア州Ｂｕｔｌｅｒ）など種々の種々の製造業者により製造されている。
【０１６８】
Ｂ．ＶＳＯＭ周辺装置
本発明の好ましい態様に使用できるデジタルカメラとしては、低照度蛍光鏡検法の用途向けに設計されたもの（例えば、ローパー・サイエンティフィック（Ｒｏｐｅｒ
Ｓｃｉｅｎｔｉｆｉｃ Ｉｎｃ．、アリゾナ州Ｔｕｃｓｏｎ）およびクウォンティテイティブ・イメージング（Ｑｕａｎｔｉｔａｔｉｖｅ Ｉｍａｇｉｎｇ Ｃｏｒｐ．、カナダ、Ｂｕｒｎａｂｙ）などの企業により製造されているもの）が含まれる。このようなカメラは、１２ビットデジタル化能力を有する科学等級のＣＣＤを有しており、かつ、約７μｍ×７ μｍのサイズである合計約１００万ピクセルを有する。このようなカメラは、Ｃ＋＋ソースコードおよびソフトウェア開発キットを含むソフトウェア開発キット、または、ホストＰＣ上でカメラ操作をソフトウェア、光学素子、電気素子、および機械素子（すなわち顕微鏡周辺装置）と統合するソフトウェアを記述するのに使用できるホスト接続キットとともに入手可能である。これらのカメラは、ホストコンピュータを介してカメラの機能をプログラムできるよう、ＰＣＩコンピュータインターフェースカードまたは直接接続の「ファイアワイア」（ＩＥＥＥ１３９４）ケーブル接続など、種々の業界標準のコンピュータインターフェースとともに提供されている。
【０１６９】
デジタルカメラに加えて他の周辺装置も本発明とともに使用でき、このような周辺装置としてはＸＹ走査ステージ、光学フィルタホイール、Ｚ軸焦点粗調整モーター、ならびに、透過光シャッターおよび表面蛍光シャッターが含まれる（例えば、ＬＵＤＬエレクトロニクス・プロダクツ（ＬＵＤＬ
Ｅｌｅｃｔｒｏｎｉｃｓ Ｐｒｏｄｕｃｔｓ， Ｌｔｄ．、ニューヨーク州Ｈａｗｔｈｏｒｎｅ）などの製造業者から入手可能な周辺装置）。種々の製造業者が他の装置を提供している。例えば、シャッター・インスツルメント（Ｓｕｔｔｅｒ
Ｉｎｓｔｒｕｍｅｎｔ Ｃｏ．、カリフォルニア州Ｎｏｖａｔｏ）は光学フィルタホイールおよびロボットマイクロマニピュレータを製造しており、シリンジポンプはハーバード・アパラタス（Ｈａｒｖａｒｄ
Ａｐｐａｒａｔｕｓ、マサチューセッツ州Ｈｏｌｌｉｓｔｏｎ）から、顕微鏡対物レンズナノポジショナ（圧電ポジショナ）はフィジック・インツルメンテ（Ｐｈｙｓｉｋ
Ｉｎｔｒｕｍｅｎｔｅ、ドイツ、Ｗａｌｄｂｒｏｎｎ）からそれぞれ入手可能である。これらの企業は、それぞれの周辺装置用に適切なプログラマブルコントローラも販売している。ハーバード・アパラタスは、種々の環境センサー（酸素、乳酸、グルコース、ｐＨなど）を操作およびホストコンピュータとインターフェースするための、ＰＬＵＧＳＹＳならびに同様の測定および制御装置も販売している。
【０１７０】
Ｃ．生物学的試薬
生物学的試薬および蛍光標識などは多数の製造業者から入手可能である。例えば、生細胞用の蛍光プローブはモレキュラー・プローブズ（Ｍｏｌｅｃｕｌａｒ Ｐｒｏｂｅｓ、オレゴン州Ｅｕｇｅｎｅ）から入手可能である。他の細胞培養培地、細胞増殖用の装置および供給品、ならびに関連する装置および供給品は、標準的な供給業者（例えばシグマ（Ｓｉｇｍａ）、フィッシャー（Ｆｉｓｈｅｒ）など）から入手可能である。
【０１７１】
Ｄ．ＶＳＯＭホストコンピュータ
本発明の好ましい態様では、１つまたは複数の使用可能なＰＣＩスロット（デジタルカメラがホストコンピュータとのインターフェースにＰＣＩスロットを必要とするか否かによる）と、標準的なシリアルポートおよびパラレルポートと、ＯＨＣＩ互換ＩＥＥＥ１３９４ポート（ファイアワイア対応デジタルカメラが必要とする場合）と、２５６ ＭＢのＲＡＭと、１つまたは複数の１８ Ｇｂｙｔｅハードドライブと、テープバックアップユニットと、イーサネットアダプタと、３２ビットカラービデオカードと、高解像度モニター（好ましくは１２８０×１０２４ピクセル）とを有し、かつ、あるバージョンのＬＩＮＵＸまたはＭｉｃｒｏｓｏｆｔ
Ｗｉｎｄｏｗｓを実行する、４５０ ＭＨｚ Ｐｅｎｔｉｕｍ ＩＩＩクラスの業界標準のＰＣを必要とする。オンラインの画像セグメント化および分析を行うために、デジタルカメラおよび種々の顕微鏡周辺装置の作動を前述のソフトウェアおよびアルゴリズムと統合させるため、標準的なＣ＋＋コンピュータプログラミングの技術が必要である。したがって、標準的なＣ＋＋プログラミングソフトウェア、およびＭｉｃｒｏｓｏｆｔ
Ｖｉｓｕａｌ Ｓｔｕｄｉｏ Ｃ＋＋などの統合ソフトウェア開発パッケージが必要である。さらに、標準的なデジタル画像取得、周辺装置の制御およびオートメーション、ならびに標準的な画像処理、プロッティング、および表示操作のためのＣ＋＋ライブラリを含むソフトウェアパッケージが望ましい。このようなパッケージとしては、「ＳＣＩＬ
Ｉｍａｇｅ」（ＴＮＯ、オランダ、Ｄｅｌｆｔ）、「ＭｅｔａＦｌｕｏｒ」および「Ｍｅｔａｍｏｒｐｈ」（ユニバーサル・イメージング・コーポレーション（Ｕｎｉｖｅｒｓａｌ
Ｉｍａｇｉｎｇ Ｃｏｒｐｏｒａｔｉｏｎ、ペンシルベニア州Ｄｏｗｎｉｎｇｔｏｗｎ））、ならびに、「Ｃｏｍｐｏｎｅｎｔ Ｗｏｒｋｓ＋＋」（ナショナル・インスツルメンツ（Ｎａｔｉｏｎａｌ
Ｉｎｓｔｒｕｍｅｎｔｓ、テキサス州Ａｕｓｔｉｎ））などがある。さらに、場合によっては、顕微鏡周辺装置の数およびＶＳＯＭシステムで制御される環境センサーの数が増加するにしたがって、ナショナル・インスツメンツが製造しているものなどの標準的なアナログ−デジタルインターフェースカードが必要となる。
【０１７２】
Ｅ．ＶＳＯＭの実施
光学プラットフォームに加えて、ＶＳＯＭの最小限の実施には以下のそれぞれが少なくとも１つ必要である：温度制御された細胞チャンバ、コンピュータ制御されたシリンジポンプ、照明シャッター、および、光学フィルタホイール。最小限で実施される実験には透過光を扱える能力は必要ではなく、蛍光鏡検法のみが必要となる。必要である唯一の微小環境制御装置は、標本容器に挿入できる温度プローブを有する、温度を３７℃に維持できる温度制御ユニットである。最小限で実施される実験では、光学プラットフォームが比較的安定であり振動がないという前提において、対物レンズのＺ軸調整は必要でない。観察の必要があるのは単一の細胞視野のみ（多視野ではない）であるため、ＸＹ位置決めステージも必要でない。しかし、本発明がこの特定の設計形式に限定されることはない。
【０１７３】
１つの態様において、細胞は、標準的な細胞インキュベータ内で標準的なプラスチック性のペトリ皿の中で標準的な技術を用いて実験前に増殖される。いくつかの態様において、細胞は、生細胞に適した核染色法により実験前に標識される。例えば、生細胞の核を染色するため、Ｈｏｅｃｈｓｔ３３３４２（Ｈ４２）が以下の方法で使用される。適切なサイズの標本容器（例えば丸型の３５ ｍｍ細胞培養皿）に２４時間付着させた細胞を、細胞インキュベータ内で１．０μｇ／ｍＬのＨ４２に約９０分間曝露する。この曝露の後、Ｈ４２および細胞増殖培地中に存在するフェノールレッド指示薬を洗い流し、フェノールレッドまたは過量のウシ胎仔血清を含まない、新鮮な培地に交換する。標本容器が細胞チャンバ内に置かれ、オペレータは適切な細胞視野に焦点を合わせ、次に画像出力をデジタルカメラに送る。コンピュータ制御のシリンジの１つには蛍光プローブを含まない適切な溶液（血清またはフェノールレッドを含まない、細胞培地など）が充填され、他のシリンジには同じ溶液と生細胞に適した蛍光プローブとが充填される。たとえば、カルセイン−ＡＭ（ＣＡＭ）は細胞株中の多剤耐性（ＭＤＲ）の存在を示すことができる適切なプローブである。最小限のＶＳＯＭ実施可能コンピュータ制御ＭＤＲアッセイでは、約１．０μＭのＣＡＭ溶液が１分あたり約０．５ ｍＬの流量で細胞チャンバ内に灌流される。この化合物は生細胞の膜を通過し、次に代謝され、蛍光性となり、かつ、細胞内に保持される。当業者には周知であるように、細胞が１つまたは複数のＭＤＲタンパク質を有するならば、その細胞はＣＡＭを細胞外に搬出する能力を有する。これらの場合、ＣＡＭの存在下において細胞が化合物を取り込む速度は低下し、かつ、ＣＡＭがチャンバから洗い流された後に細胞が保持するＣＡＭの量は少なくなる。
【０１７４】
ＶＳＯＭシステムはＣＡＭ溶液の灌流を開始し、かつ、核内のＨ４２の青い核蛍光信号を用いて生細胞の視野を連続的にセグメント化する。このように、核を描く輪郭が青チャネルで取得される。核領域におけるＣＡＭの平均蛍光強度を計算するため、これらの同じ輪郭が緑チャネルでも使用される。ＣＡＭは細胞全体に均一に分布することが多いため、核領域に加えて細胞質領域を検出する必要はない。適切な間隔（例えば６０秒ごと）で画像が取得されると同時に、このように緑チャネルにおける平均細胞内蛍光強度（ＭＩ）が各細胞の核領域について観察される。
【０１７５】
次に、下記のような方法で細胞チャンバ内のＣＡＭ量を調整するため、コンピュータアルゴリズムが以下のサーボループオペレーションを実行する。各画像取得点について全細胞の平均ＭＩ（前述に定義のとおり）がモニターされ、次に以下が実行される：
ａ）視野内の全細胞の平均ＭＩがユーザー定義の閾値に達すると、ＣＡＭの入ったシリンジが遮断され、かつ、ＣＡＭの入っていないシリンジがオンになる；または、
ｂ）視野内の全細胞の平均ＭＩが、（ａ）で達した最大平均ＭＩからユーザー定義のパーセンテージ分減少すると、ＣＡＭの入っていないシリンジが遮断される；
ｃ）実験終了。
【０１７６】
このような最小限の実験において、ＣＡＭの取込みおよび保持の個々の率が細胞ごとに取得される。システムは次に、実験の最後に、視野内においてＭＤＲを示した細胞の数（またはパーセンテージ）を報告することができる。さらに、このような細胞はこれによって同定され、かつ、ＭＤＲを変化させる化合物を試験する次の試験に供することができる。しかし、本発明がこの特定のシステム、形式、細胞などに限定されることはない。事実、本発明は実験デザインの面においてユーザーに最大限の柔軟性を提供する。
【０１７７】
実施例３
ＶＳＯＭ実験
本実施例では、実施例２の説明に以下に記載の変更点を加えて実施したＶＳＯＭ実験について説明する。
【０１７８】
使用した光学プラットフォームは、透過光鏡検法（位相差法およびＤＩＣ法）および多色蛍光鏡検法用の装備を有するＴＶ倒立顕微鏡、Ｚｅｉｓｓ Ａｘｉｏｖｅｒｔ１３５ Ｈ／ＤＩＣである。顕微鏡には、コンピュータ制御のＸＹ走査ステージ、Ｚ軸ステップモーター、および６位置フィルタホイール（ＬＵＤＬエレクトロニック・プロダクツ（ＬＵＤＬ
Ｅｌｅｃｔｒｏｎｉｃ Ｐｒｏｄｕｃｔｓ， Ｌｔｄ．、ニューヨーク州Ｈａｗｔｈｏｒｎｅ））を装備した。これらの研究では、Ｋｏｄａｋ ＫＡＦ−１４００ ＣＣＤチップ（１３１７×１０３５ピクセル、７×７ミクロンピクセルサイズ）を有する１２ビットのＸｉｌｌｉｘ
ＣＣＤカメラ（キシリックス・テクノロジーズ（Ｘｉｌｌｉｘ Ｔｅｃｈｎｏｌｏｇｉｅｓ、ブリティッシュコロンビア州Ｖａｎｃｏｕｖｅｒ））を使用した。このカメラは８ＭＨｚ（すなわち、１セットあたりフルサイズ画像約４枚）の読み出し速度を有する。ホストコンピュータはマルチタスクＵＮＩＸワークステーションであるＳｐａｒｃｓｔａｔｉｏｎ
Ｕｌｔｒａ １とし、コンピュータカメラから出力された画像をホストコンピュータに直接読み取らせた。
【０１７９】
さらに、ペルティエ温度制御マイクロ灌流チャンバ（ＴＣ−２０２二極式温度制御装置を有するＰＤＭＩ−２開放チャンバ、ハーバード・アパラタス／メディカル・システムズ・リサーチ・プロダクツ（Ｈａｒｖａｒｄ
Ａｐｐａｒａｔｕｓ／Ｍｅｄｉｃａｌ Ｓｙｓｔｅｍｓ Ｒｅｓｅａｒｃｈ Ｐｒｏｄｕｃｔｓ、マサチューセッツ州Ｈｏｌｌｉｓｔｏｎ））およびデュアルシリンジポンプ（Ｐｕｍｐ−３３、ハーバード・アパラタス）を使用した。これらのシリンジは、同じサイズである必要はなく、かつ、灌流速度が独立に制御できるよう各シリンジがそれぞれコンピュータ制御のブロックを有する。「浴槽」タイプのサーミスタ（ＢＳＣ−Ｔ３、計３６Ｋオーム）をＰＤＭＩ−２とともに使用した。
【０１８０】
いくつかの好ましい態様において顕微鏡の設計が倒立式であることによって、生細胞の灌流が可能となり、不動化された細胞が下から撮像された。本発明のいくつかの態様では油浸対物レンズが使用できる。また、細胞は、いくつかの態様では３５ｍｍペトリ皿の中で、他の態様では円形のカバーガラス上で増殖され、下から観察される。この倒立顕微鏡の構成により、ユーザーは蛍光から透過光検出（位相差またはＤＩＣ）に切り替えることができ、かつ、細胞の位置および形態を確認することができる。これはコンピュータ制御の透過光シャッターにより実現される。フィルタホイールにあるコンピュータ制御の表面蛍光シャッターも使用可能である（ＬＵＤＬ）。これらの実験において、システムの自動視覚サーボ制御オペレーション（すなわち、ポンプ、照明シャッター、フィルタホイール、およびＸｉｌｌｉｘカメラの制御）はプログラム「ＡＤＲ」で実施した。あらかじめプログラムされたレシピを使用して実施したＶＳＯＭ実験ではプログラム「ＶＸ５」を使用した。
【０１８１】
Ａ．ＡＤＲ視覚サーボ制御実験
ＭＣＦ−７ ＷＴＣ細胞を３５ ｍｍ細胞培養皿で増殖させ、Ｈ４２で前染色した。Ｈ４２とともに約１時間プレインキュベーションした後、細胞をＤＰＢＳＧＰですすぎ、温めた顕微鏡チャンバに入れた。ＡＤＲプログラムは以下の要領で問題なく動作した。この実験中、指定した一定の間隔で３枚のデジタル画像（３６０ｎｍ励起、４９０ ｎｍ励起、および透過光、明視野）が取得され、かつ、次の撮像間隔より前に、Ｈ４２で染色されたすべての核が検出およびセグメント化された。青チャネルで描出された各核の輪郭を用いて、平均細胞蛍光強度の値が緑チャネルから算出された（ＭＩ）。ＤＢＰＳＧＰ溶液のみ（シリンジ＃１）による最初の灌流を１．０ｍＬ／分で８．３分間実施し（ユーザーの指定による）、次にシリンジ＃１が停止された。次に、１ μＭのＣＡＭを含むＤＢＰＳＧＰ溶液（シリンジ＃２）の灌流を０．５ｍＬ／分で開始した。全細胞の平均ＭＩが指定の閾値に達したときに、正常な視覚サーボ制御によりシリンジ＃２が停止された。細胞が十分量のＣＡＭを蓄積しかつシステムが十分量の細胞内ＣＡＭを（核領域で）検出し、システムが正常にシリンジ＃２を停止させかつシリンジ＃１を再始動させるまでに、１３．９分を要した。１１．４分後（ユーザーの指定による）に、システムは再度シリンジ＃１を停止させた。システムは引き続きチャンバ内の細胞をモニターし（ユーザーの指定による）、流れがない状態でさらに２４．９分間が経過した。次に、実験終了時にシステムは画像取得を停止した。
【０１８２】
実施例４
ＭＤＲアッセイ−試験キットおよびＶＳＯＭ実験
本明細書に記載のとおり、ＶＳＯＭ実験においてＭＤＲアッセイβ試験キットの試験も行った。図１１にこのＶＳＯＭ実験の略図を示す。各ポンプは、あらかじめプログラムされたレシピに従って、または個々の細胞反応のリアルタイム分析に基づいて、指定された流量でオンおよびオフになる。使用したＣＡＭ、Ｖ、ＭＫの濃度を図中に示した。ＣＡＭへの３回の曝露はそれぞれ９００秒間（２０分間）であり、したがってＣＡＭへの総曝露時間は６０分間であった。本図の下部の黒い窓は視野内の全細胞の反応の平均を示す。この平均反応プロットはＶＳＯＭ実験中コンピュータ画面に表示され、かつ、１つまたは複数のチャネルのデジタルイメージも画面に表示される（通常、表示されるチャネルの１つは透過光である）。
【０１８３】
実験中、プロットおよび画像は継続的に再描画および更新される。ＭＣＦ−７（ＤＳ、実行（ｒｕｎ）＃１）およびＭＣＦ−７ ＡＤＲ（ＭＤＲ、実行＃２）について記録された２つのプロットにおいて、ＤＳ細胞のカルセインの蓄積量（２４０± ８２、Ｎ＝３３６）はＭＤＲ細胞の蓄積量（１０３ ± ８０、Ｎ＝１６１）より多いことが見出された。このことは、多数の細胞の平均を表すマルチウェルプレート実験で観察される、ＤＳ細胞とＭＤＲ細胞とのカルセイン蓄積の相対的な差異と一致する（前述の表１参照：１６５８／４８２２＝０．３４、１９６５／６５９８＝０．３０、ＶＳＯＭ：１０３／２４０＝０．４３）。しかし、ＶＳＯＭ実験では各細胞の平均細胞蛍光がモニターされ、かつ、平均値からの逸脱の検出が容易であった。例えば、複数の細胞が生存率および原形質膜完全性を失い始めた。このような細胞はカルセインを保持することができず、カルセインが漏出してその結果蛍光信号が低下する。これは、前述のＤＭＳＯの毒性に関連した影響である可能性がある。しかし、本発明を使用する上で機序の理解は必要ではなく、かつ、本発明が特定の機序に限定されることはない。それでもなお、単一の細胞の死がみとめられた時点で容易に曝露／灌流を終了させる能力を現在のＶＳＯＭシステムが有することは注意を要する。
【０１８４】
図１３は、本実施例の実行＃２の結果から、選択した４つの細胞の細胞反応を示したものである。これらの反応曲線はβ試験キットの基礎をなす原理を示すよい例である。反応曲線のうち２つでは（図１３）、シリンジ＃５の作動中に（図１１）ベラパミル（Ｖ）に曝露されるまで細胞はカルセインを蓄積しない。ＰｇｐポンプはＭＫ−５７１（ＭＫ）で阻害されないことから、これらの細胞はＰｇｐポンプのみを発現していると解釈することができる。ＤＳ細胞の母集団中にこれらの細胞のいずれも存在しないことは明らかである（図１２）。さらに、ＣＡＭ＋Ｖ曝露後、１つの細胞は、その平均細胞蛍光が、ＤＳ細胞で観察される平均細胞反応の一方の標準偏差の範囲内に達するまで急速にカルセインを蓄積した（図１２）。他のすべての要因が同じであると仮定すると、第二の細胞はＰｇｐ分子の含有量が少ないと予測できる。残る２つの曲線（図１３）は、ＣＡＭ（のみ）およびＣＡＭ＋ＭＫの曝露中に別の反応を示しており、これはさらなる分析を要する。観察された曲線は、内在化したカルセインを排出する能力をもたない２種類のポンプのみの存在または不在の観点からは説明できない。しかし、本発明を使用する上で機序の理解は必要ではなく、かつ、本発明が特定の機序に限定されることはない。これは、フローサイトメトリーアッセイおよびマルチウェルプレートアッセイでは得られない種類の詳細な生理学的情報である。対照的に、本明細書に示したＶＳＯＭ技術はこれらの能力を有している。
【０１８５】
例えば、カルセインの蓄積および保持に関する詳細な率の情報がＶＳＯＭ実験全体を通して利用可能である。いくつかの反応曲線について、３回のＣＡＭ灌流中のカルセイン蓄積率を計算することができる。さらに、ＢＵＦＦ灌流中のカルセイン保持率も計算することができる。カルセイン蓄積は１５００秒、３５００秒、および５５００秒の時点で生じたことが見出され、かつ、良好なカルセイン保持に対応するプラトーは２５００秒、４５００秒、および６５００秒のインターバルで観察された。前述のとおり、いくつかのＭＤＲポンプはカルセインが細胞内に内在化された後にこれを排出する能力をもっている。内在化されたカルセインを排出するＤＳ細胞の能力は、ＢＵＦＦ灌流に対応する２５００秒の期間中に顕著である（図１２参照）。しかし、２つのＭＤＲ細胞（図１３）はこの能力をもたない。
【０１８６】
さらに、前述の教示を使用すると、コンピュータが判断を行うあらかじめプログラムされたマニピュレーションは視覚的サーボ制御オペレーションに適している。
【０１８７】
実施例５
修飾試薬を用いたデジタル撮像蛍光法
ＶＳＯＭを使用して、デジタル撮像蛍光顕微鏡を自動化することができる。例えば、修飾試薬ＭＫ−５７１（１０ μＭ）またはベラパミル（５０ μＭ）を含むカルセイン−ＡＭ（ＣＡＭ、０．２５μＭ）をマイクロ灌流チャンバに灌流させることができる。ＭＫ−５７１は膜貫通タンパク質ＭＲＰに対する特異的な阻害剤であり、一方、ベラパミルはＭＲＰを阻害するとともに膜貫通タンパク質ＰｇＰも阻害する。これらの膜貫通タンパク質はいずれも外来性の化合物を排出し、これらタンパク質は多剤耐性に関与していると考えられている。これら２つの異なるタンパク質をコードしているのは別々の遺伝子である。これらタンパク質の薬剤交差耐性プロフィール（スペクトル）は、重複するものの完全に同じではない。さらに、前述のとおり、これらタンパク質は種々の阻害剤に対して異なる感受性を有する。
【０１８８】
Ｈｏｅｃｈｓｔ ３３３４２で標識したＭＣＦ−７ＡＤＲ細胞を３５℃で観察し、細胞ごとの平均蛍光強度（ＭＩ、カルセイン）を求めた。細胞ごとの平均蛍光強度（ＭＩ、カルセイン）はリアルタイムで計算され、シリンジポンプのソフトウェアオペレーションはこれらの計算結果に基づいて実行される（すなわち視覚的サーボ制御）。これらの実験で使用されるプロトコルは、前述のＭＤＲベータ試験キット（モレキュラー・プローブズ）を基礎にしている。画像は以下の３つの灌流インターバル中に取得される：（１）０〜２１００秒、ＣＡＭ；（２）８６００〜１１，０００秒、ＣＡＭ＋ＭＫ−７５１；および（３）１４，０００〜１６，０００秒、ＣＡＭ＋ベラパミル。典型的には、３番目の灌流インターバルまでにすべての細胞が反応する。ＭＲＰおよび／またはＰｇＰを発現している部分母集団は、以下のように、ＭＩ＞ ２００となる反応に基づいて推論される：（ａ）１で反応（ＰｇＰおよびＭＫ−５７１のいずれも発現なし）、（ｂ）２まで反応しない（ＭＲＰ）、および（ｃ）１または２、のいずれでも反応しない（ＰｇＰのみ）。インターバル１の間はＰｇＰおよびＭＫ−５７１のいずれの発現も観察されなかったが、インターバル２ではＭＲＰの発現が観察された。ＭＫ−５７１によるＭＲＰ阻害のため、３０個の細胞においてカルセイン蓄積によるＭＩの増加が２００を上回った。３枚のデジタル画像（各チャネルは６０秒ごとに取得された）。例えば、細胞＃１０のＭＩは、この２４００秒の灌流インターバル中に３０から７６０に増加した。
【０１８９】
実施例６
生細胞用のＢｒｄＵ増殖アッセイ
本実施例では、生細胞用のＢｒｄＵ増殖アッセイシステムについて説明する。特定の細胞反応を特定の生物学的エンドポイントに相関付けできるならば、任意のＶＳＯＭ予測アッセイの予測力が向上する。アポトーシス用の蛍光アッセイは複数存在するが、ＤＮＡ合成または細胞増殖の定量化に使用できるＢｒｄＵ取込みなどの重要なパラメータを測定できる生細胞用の蛍光アッセイはほとんどない。これらの実験において、ＢｒｄＵ取込み量を定量化することによって生細胞の増殖状態を検出するための画像比演算技術が開発された。このアッセイのためのプロトコルは、発表されているＤｒ．Ｐａｕｌ Ｙａｓｗｅｎ（Ｓｔａｍｐｆｅｒら、Ｅｘｐ． Ｃｅｌｌ Ｒｅｓ．、２０８：１７５−１８８ ［１９９３］）のプロトコルをもとに適合させたものである。細胞株および培地は同じものを用いた。しかし、ＤＮＡ合成アッセイ用として、［^３Ｈ］−チミジンの代わりに５−ブロモ−２’−デオキシ−ウリジン（ＢｒｄＵ）を使用した。これらの実験は、５−ブロモ−２’−デオキシ−ウリジン標識および検出キット（Ｎｏ．１２９６７３６、ベーリンガー・マンハイム（Ｂｏｅｈｒｉｎｇｅｒ Ｍａｎｎｈｅｉｍ））を使用して行った。ＶＳＯＭ実験の後、細胞を固定し、かつ、従来の間接免疫蛍光アッセイを用いてデジタル画像比演算に基づいてＶＳＯＭの結果を確認した。
【０１９０】
増殖するよう刺激を与えた細胞は、細胞分裂の準備においてＤＮＡの相補鎖を通常の２倍合成する。活発に増殖するこれらの細胞は、細胞外の培地からＡ、Ｔ、Ｃ、およびＧのヌクレオチドを取り込み、これらをＤＮＡの合成に使うことができる。しかし、ＢｒｄＵの存在下ではＴの代わりにＢｕｄＵがＤＮＡ合成に使われる。したがって、細胞の母集団を短時間（例えば１時間）ＢｒｄＵに曝露すると、その１時間の曝露中にＤＮＡを合成した細胞の部分母集団ではほとんどのＢｒｄＵがその核のＤＮＡに組み込まれる。これは「パルスラベル」実験と呼ばれる。この実験は、Ｈｏｅｃｈｓｔ３３３４２（Ｈ４２）およびＳｙｔｏ １６（Ｓ１６）の蛍光発光強度の比が、ＢｒｄＵ取込み量を細胞単位で決定する上で有用か否かを決定することを目的に実施した。
【０１９１】
Ｈ４２およびＳ１６は生細胞のまたは固定された細胞の核を染色する。Ｈ４２の蛍光発光はＤＮＡにＢｒｄＵが組み込まれていると消失（減少）する。しかし、Ｓ１６の蛍光発光は組み込まれたＢｒｄＵの存在による影響を受けない。したがって、Ｈ４２／Ｓ１６蛍光強度の比（デジタル画像を用いてピクセルごとに計算される）は、各細胞のＤＮＡに組み込まれたＢｒｄＵの量に比例すると考えられる。
【０１９２】
これを決定するため、これらの実験では１８４Ｂ５細胞の３つの培養皿を使用した。最初の４８時間、２つの培養皿には無血清のＭＥＢＭ（乳腺上皮細胞基本培地、ＣＣ−３１５１、クロンティックス（Ｃｌｏｎｅｔｉｃｓ））を使用し、３番目の培養皿には無血清のＭＥＧＭ（乳腺上皮細胞増殖培地、ＭＥＧＭ、ＣＣ−３０５１、クロンティックス（Ｃｌｏｎｅｔｉｃｓ））を使用した。ＭＥＧＭはウシ下垂体抽出物（ＢＰＥ）、ヒドロコルチゾン、ヒト上皮成長因子（ＥＧＦ）、およびインシュリンを含むが、ＭＥＢＭはこれらを含まない。したがって、最初の４８時間、培養皿３には増殖に一般的に必要な因子がすべて存在したが、培養皿１および２にはこれらが存在しなかった。ＭＥＢＭは細胞を基礎代謝レベルで支持するのみであり、細胞の付着または増殖を支持するよう意図されたものではない。４８時間後、培養皿２にはＭＥＧＭを添加し（この２段階処理を「ＭＥＢＭ＋ＥＧＦ」と呼ぶ）、培養皿３にもＭＥＧＭを添加したが、培養皿１にはＭＥＧＭ−ｈＥＧＦ（ＢＰＥ、ヒドロコルチゾン、およびインシュリンは存在）を添加した。培養皿１の処理を「ＭＥＢＭ−ＥＧＦ」と呼ぶ。
【０１９３】
Ｓｔａｍｐｆｅｒらの図４で示されているように、１８４Ｂ５細胞では供給から約１８時間後にＤＮＡ合成がピークに達し、ＤＮＡ合成量はＭＥＧＭ細胞が最大となり、ＭＥＢＭ＋ＥＧＦ細胞ではそれより少なかった。ＭＥＢＭ−ＥＧＦに維持した細胞ではＤＮＡ合成がほとんど認められなかった。これらの結果は［^３Ｈ］−チミジンを用いて得られたものである。本発明の開発中に実施した実験では、ＢｒｄＵによるパルスラベルを１時間行い、次に生細胞をＨ４２およびＳ１６で二重染色した。細胞のデジタル画像を３６０ｎｍ励起（Ｈ４２信号）で１枚取得し、かつ、４９０ ｎｍ励起（Ｓ１６信号）で第二の画像を取得した。これらの画像を処理し、ピクセルごとに比を計算した。次に、分析の便宜のため、得られた比の値をカラーコード化した。これらの結果は、従来の［^３Ｈ］−チミジン組込みアッセイで得られたものと一致していた。ＢｒｄＵ組込みの程度はＭＥＧＭ細胞（培養皿３）が最も大きく、次に大きかったのがＭＥＢＭ＋ＥＧＦ細胞（培養皿２）であり、ＭＥＢＭ−ＥＧＦ細胞（培養皿１）ではＢｒｄＵ組込みがほとんどみとめられなかった。
【０１９４】
これらの観察結果をさらに確認するため、７０％エタノールおよび５０ ｍＭグリシンで、ｐＨ ２．０、−３０℃で細胞を固定し、間接免疫蛍光染色を行った。この染色後、抗ＢｒｄＵ抗体で標識された固定細胞にＢｒｄＵ消失を示すＨ４２蛍光の低下もみとめられるか否かを決定するため、Ｈ４２およびＳ１６で細胞を再染色した。実際にこの影響を示す証拠がいくつか存在した。ＭＥＧＭ細胞が最大量のＢｒｄＵ染色を示した。同細胞を青チャネルで調べたところ、ＢｕｄＵで染色される細胞はＨ４２蛍光信号が低いという一般的な傾向がみとめられた。ＢｒｄＵ取込みを定量化する方法は２つある。第一は抗ＢｒｄＵ抗体検出であり、これは赤色の蛍光の強度で表される。第二は、青色の核蛍光強度（Ｈ４２）を緑色の核蛍光強度（Ｓ１６）で除した比を細胞ごとに計算するものである（すなわち画像比演算）。ＢｕｄＵ取込みによってＨ４２は消光するがＳ１６は消光しないため、赤色を示す核は青チャネルにおいてＨ４２蛍光の低下を示すと考えられる。
【０１９５】
実施例７
ｍＲＮＡ転写の画像化
本実施例では、ＶＳＯＭ実験の長所を示す実験について説明する。このような実験は、アンチセンス化合物、送達システム、および信号増幅法の開発の補助として、組織培養中のｍＲＮＡ転写をリアルタイムで画像化する手段を提供すると予測される。リアルタイム医用画像技術に適した放射線標識アンチセンス化合物を提供する実験についても説明する。
【０１９６】
これらのＶＳＯＭ実験では、ｂｃｌ−２タンパク質をコードしているｍＲＮＡとともに、モデル系としてヒト乳癌細胞株ＢＴ−４７４が使用される。個々の細胞内の蛍光標識したアンチセンス化合物が追跡され、細胞内の適切な標的ｍＲＮＡに誘導され、かつ、個々の細胞が予測された方式で影響されているか否かが決定される。デジタル画像蛍光鏡検法、新規性のある計算技術、およびバイオインフォマティクスツールの使用により１つのチャネルで蛍光アンチセンス化合物が追跡され、別のチャネルでは他の細胞反応（例えばアポトーシス）がモニターされる。ｂｃｌ−２のｍＲＮＡレベルでの発現が、時間、細胞外ストレス、およびアンチセンス化合物の構造の関数として画像化される。これらの観察結果は各細胞内のｂｃｌ−２タンパク質レベルおよび各細胞の最終運命と相関付けられる。したがって、これらの実験は、ＰＥＴで遺伝子発現を画像化するためのアンチセンス化合物のインテリジェント設計に有用な、特定のアンチセンス化合物の生物物理学的特性に関する詳細な情報を提供する。事実、本発明のインビトロ技術により、インビボ研究（例えば、異種腫瘍細胞株ＢＴ−４７４を移植したヌードマウスにおける研究）の場合よりも多くの化合物修飾を試験することが可能となる。
【０１９７】
いくつかの態様において、ＶＳＯＭ実験とインビボＰＥＴ撮影試験の両方が可能となるよう、同じ配列だが５’末端を修飾したＡＳ−ＯＤＮ化合物を使用した。ＶＳＯＭ実験では蛍光色素を使用し、ＰＥＴ撮影試験では２つのフッ化放射性成分の１つを使用した。いずれの場合も、蛍光色素および放射性標識の付着を容易にするため５’末端にヘキシルアミンリンカーを結合させた。追加の実験では、送達ビークルとしてリポソームおよびイムノリポソームを用いた。いくつかの実験では、放射線標識試験で使用したフッ化成分の非蛍光、非放射性のバージョンを生物学的エンドポイントの基礎とした。この方式により、ＡＳ−ＯＤＮの５’末端における種々の置換基の結果を決定した。さらに、β−ガラクトシダーゼプロトコルで得られる信号増幅の量を本発明とともに使用することができる。他の態様ではペプチド核酸が使用される。ペプチド核酸（ＰＮＡ）は、デオキシリボース骨格全体がポリアミド（ペプチド）鎖で置換された、修飾オリゴヌクレオチドである（Ｇｅｗｉｒｔｚら、Ｂｌｏｏｄ９２：３６ ［１９９８］； ならびに、ＴｅｍｓａｍａｎｉおよびＧｕｉｎｏｔ、Ｂｉｏｔｅｃｈｎｏｌ． Ａｐｐｌ． Ｂｉｏｃｈｅｍ．、２６：６５−７１［１９９７］）。
【０１９８】
１つの態様では以下のＡＳＤ−ＯＤＮ配列を使用した。

【０１９９】
ｂｃｌ−２のオープンリーディングフレームの最初の６コドンに対するアンチセンス配列を有する、配列番号：１に示した１８−ｍｅｒの完全なホスホロチオエートは、重症免疫不全マウスに移植したＤｏＨＨ２リンパ腫に対する有効性が認められている（Ｒａｙｎａｕｄら、［１９９７］）。１５−ｍｅｒの２つのＰＮＡ−ＯＤＮ（配列番号：２および３）は無細胞系でのインビトロ評価が行われている（Ｍｏｌｏｇｎｉら、［１９９９］）。特に興味深いのは、ｍＲＮＡ転写が完全に阻害されるのは両方のＰＮＡ−ＯＤＮが同時に存在する場合のみであるということである。
【０２００】
蛍光標識したＰＴ−Ｇ３１３９、ＰＮＡ−１、およびＰＮＡ−４を市販の供給源から入手した（例えばＧｅｎｓｅｔ、ｈｔｔｐ：／／ｗｗｗ．ｇｅｎｓｅｔｏｌｉｇｏｓ．ｃｏｍ）。これらＡＳ−ＯＤＮの修飾物であり骨格がＰＤ、ＰＴ、ＭＰ、またはＰＮＡの骨格結合で完全に構成されるもの、ならびに、骨格結合に種々の置換および混合を有する化合物も使用した。
【０２０１】
最初の実験を行い、一連の単一細胞の反応としてデータベースに記録した。前述のように骨格結合に種々の修飾を有する、蛍光標識したＡＳ−ＯＤＮのセットを使用した。ベースラインとなるＶＳＯＭ実行のセットを行った。これらの実行において、細胞は顕微鏡カバーグラス上で増殖させ、かつ、細胞観察用に６３×ＮＡ１．３油浸対物レンズを使用した。細胞を温度管理した微小灌流チャンバに入れた。細胞質区画、核質区画、およびミトコンドリア区画の最初のセグメント化を以下の要領で行った。
【０２０２】
カルシウム、マグネシウム、グルコース、およびピルビン酸塩を含むダルベッコのリン酸緩衝生理食塩水（Ｄ−ＰＢＳ）に溶解した蛍光化合物ＴＭＲＥ（テトラメチル・ローダミン・エチルエステル；モレキュラー・プローブズ）をコンピュータ制御のシリンジポンプにより細胞チャンバに注入する。化合物ＴＭＲＥは、Ｎｅｒｓｔの式（Ｆａｒｋａｓら、前述）によってミトコンドリアの膜電位を決定するために使用される、再配分型のカチオン染料である。ミトコンドリア染料であるローダミン１２３と異なり、ＴＭＲＥは迅速かつ可逆的にミトコンドリア内で平衡し、かつ、程度はより少ないものの細胞質でも平衡する。核質区画はＴＭＲＥで染色されない。
【０２０３】
次に、緩衝液のみが入ったコンピュータ制御のシリンジを使用して細胞からＴＭＲＥを洗い流す。ＴＭＲＥは細胞から容易に洗い流すことができる。次に、別のコンピュータ制御シリンジを使用して、蛍光標識ＡＳ−ＯＤＮを微小灌流チャンバに注入する。ＡＳ−ＯＤＮが発する蛍光信号が可視状態になるときの濃度が記録され、かつ、ＡＳ−ＯＤＮの灌流が自動的に停止される。これがＶＳＯＭ実行の前半の終了である。
【０２０４】
ＶＳＯＭオペレーションを確認するための再生ができるよう、すべてのデジタル画像がデータベースに追加される。このように、多チャネルデジタル画像、実験パラメータ、および、コンピュータ制御の化合物灌流に対する単一細胞の反応（６３×で細胞５〜１０個）がデータベースの一部となる。
【０２０５】
最初のＶＳＯＭ実行の後半では、蛍光標識ＡＳ−ＯＤＮの蛍光分布パターンと、細胞小器官特異的蛍光染料（モレキュラー・プローブズ）を用いて同じ細胞種について以前に取得された蛍光分布パターンまたは蛍光ＡＳ−ＯＤＮと細胞小器官特異的蛍光染料とで二重染色した現在観察中の細胞について別の蛍光チャネルで取得した現在の蛍光パターンとの比較が行われる。ＶＳＯＭは、細胞内のＡＳ−ＯＤＮの位置を決定し終えると、適切なシリンジから適切な化合物を添加して細胞内のＡＳ−ＯＤＮ分布に対する影響を観察する。
【０２０６】
蛍光標識ＡＳ−ＯＤＮのリソソーム隔離が生じる場合は、リソソームを選択的に透過化することが知られている化合物が、コンピュータ制御の灌流シリンジにより制御された流量でチャンバ内に灌流される。モネンシンプロトンイオノフォアナトリウムなど複数の化合物が、ヒト乳癌細胞の酸性小胞内に捕捉された蛍光化合物を解放することが明らかになっている（Ｓｃｈｉｎｄｌｅｒら、［１９９６］を参照）。蛍光標識ＡＳ−ＯＤＮを解放するのに必要な（例えば）モネンシンの量がデータベースに記録される。隣接区画（例えば核）への再分布に要した時間およびその結果生じた核内の染色パターンも記録される。
【０２０７】
核からＡＳ−ＯＤＮを除去するためのＶＳＯＭ摂動は、当業者には周知である「クランピング」技術を使用して実施される。これらの技術により、ユーザーは、単純に周囲の細胞外培地のイオン（例えばカリウム）の濃度を変化させることにより細胞内のカリウム濃度を変化させることができる（Ｎｅｇｕｌｅｓｃｕら、前述）。これと同じ方法を使用することによって細胞内のカルシウムレベルを変化させることができる。ＶＳＯＭシステムは、核の塩基性タンパク質に対するＡＳ−ＯＤＮの結合に干渉するため、細胞内のこのようなイオンの濃度を漸増させる。この場合も、核環境から化合物を解放させるのに必要な濃度およびその他の環境条件が記録される。この方法が成功しない場合は、オンラインの低張細胞溶解が実行され、かつ、以後の実験ではより厳しい環境条件が使用される。
【０２０８】
いくつかの態様において、ＶＳＯＭは、細胞膜に結合したＡＳ−ＯＤＮを解放するため、非生理学的なｐＨの緩衝液によるすすぎまたはトリプシン含有緩衝液によるすすぎによって摂動を与える。この場合も、蛍光標識ＡＳ−ＯＤＮの解放に必要な摂動の大きさに関する定量的データが細胞ごとに記録され、かつ、現在使用中のＡＳ−ＯＤＮの化学構造と相関付けられる。
【０２０９】
好ましい態様において、（蛍光性および非蛍光性の）ＡＳ−ＯＤＮの生理学的影響が複数の方法で評価される。多くの場合はアポトーシス誘発性細胞外ストレス（例えば、紫外光、抗癌剤、カルシウムイオノフォアなど）が印加される。ＡＳ−ＯＤＮによるＢＴ−４７４細胞中のｂｃｌ−２タンパク質量の低減が成功した場合は、ｂｃｌ−２タンパク質の減少およびこれらの細胞のアポトーシスに対する抵抗能力が観察される。最も直接的な方法では、ＶＳＯＭ実験の後に細胞が固定され、ｂｃｌ−２タンパク質に対する抗体を使用した間接免疫蛍光分析が行われ、続いて、同じ細胞視野に戻ってｂｃｌ−２タンパク質の定量が行われる（これは、細胞ごとに観察される蛍光信号と比例する）。しかし、本発明が特定の段階のセットに限定されることはなく、当業者には他の方法も本発明との関連で使用するのに適していることが理解されるものと思われる。
【０２１０】
当業者に周知の他の多くのアポトーシスマーカーも本発明とともに使用可能である。これらの方法には形態学的アッセイシステムおよび蛍光アッセイシステムが含まれる。例えば、ＶＳＯＭ実験で収集されるチャネルの１つは透過光チャネルであり、このチャネルを介して膜のブレブを観察することができる。また、アポトーシスに特徴的な核断片化を明らかにできる、種々の抗タンパク質（例えばアネキシンＶ）抗体およびＤＮＡ染料も存在する。さらに、すでに詳述したように、細胞膜完全性（例えば細胞の生存率）を示すカルセイン−ＡＭなどの染料も、ＶＳＯＭ実験中に蛍光チャネルの１つで使用するのに適している。さらに、ＬＤＳ−７５１で染色した１８５ｂ５細胞についてすでに詳述したように、アポトーシスの早期に生じるミトコンドリアの膜電位の消失もアポトーシスの指標として有用である。
【０２１１】
以上より、本発明が細胞種間の差異に関する知識ベースの発見および最適化のための改良された方法およびシステムを提供することは明らかであると思われる。特に本発明は視覚的サーボ制御光学顕微鏡および分析方法を提供する。本発明は、数百の個々の生細胞を経時的に詳しくモニターする手段；多チャネルにおける動的生理学的反応の定量化；リアルタイムのデジタル画像セグメント化および分析；インテリジェントなかつ繰り返しの、コンピュータ印加による細胞ストレスおよび細胞刺激；ならびに、長期的な研究および観察において同じ細胞視野に戻ることができる能力を提供する。本発明はさらに、特定の細胞部分母集団用に培養条件を最適化する手段を提供する。
【０２１２】
本明細書で言及した発表物および特許はすべて参照として本明細書に組み入れられる。当業者には、本発明の範囲および精神から逸脱することなく、本明細書に記載された本発明の方法およびシステムに種々の改変および変形を行い得ることが明らかであると思われる。特定の好ましい態様に関して本発明を説明したが、特許請求の範囲で請求されている本発明がこのような特定の態様に不当に限定されることはないと理解されるべきである。事実、本発明を実行するための本明細書に記載の方式に対する、当業者に明白である種々の改変は、本明細書に記載の特許請求の範囲に含まれるものと意図される。
【図面の簡単な説明】
【図１】本発明のマイクロプロセッサ／コンピュータ、ロボット、および光学システム、ならびにシステムのその他の構成要素間の関係を示す図である。
【図２】本発明の代替的態様を示す別の図である。
【図３】Ｂｌｏｂ ＭａｎａｇｅｒおよびＩｎｓｔｒｕｍｅｎｔ Ｍａｎａｇｅｒの主要なオブジェクトの関係を示す図である。追加の機器はそれぞれＣｏｎｔｒｏｌＳｏｕｒｃｅおよびＢｌｏｂＳｏｕｒｃｅ用のプラグインを必要とする。ＣｏｎｔｒｏｌＳｏｕｒｃｅおよびＢｌｏｂＳｏｕｒｃｅの詳細はＩＤＬレベルで隠されている。
【図４Ａおよび図４Ｂ】ＶＳＯＭ実験の１つの態様を示す略図である。
【図５】多視野（ＭＦ）ＶＳＯＭ実験の１つの態様を示す略図である。
【図６】２つのデータモデルを提供する。パネルＡはバイオインフォマティクスシステム用の完全データモデルであり、パネルＢはインビボ処置用の詳細な内訳（ｂｒｅａｋｄｏｗｎ）である。
【図７】分割の過程を示す。パネルＡは、組織および細胞培養物から個々の核を抽出するためのプロトコルを提供する。パネルＢは、隣接する２つの核を用いた重心の変換の様子を示す。
【図８】元の細胞を示す写真である。
【図９】ＤＡＰＩにおける分割の結果を示す。
【図１０】本発明の知識ベースのＶＳＯＭの１つの態様における種々の構成要素の相互作用を示す図である。この図において、イタリック体の項目は全体のシステムアーキテクチャ（ａｒｃｈｉｔｅｃｔｕｒｅ）で定義されるサービスを指す。
【図１１】本明細書に記載されたＶＳＯＭ実験（実施例５を参照）のいくつかで使用される、シリンジ、シリンジ内の溶液中の組成物、流速、および灌流間隔に関する詳細情報の図である。両方の実行（ｒｕｎ）においてポンプスケジュールを用いた。２回の実行を（同じ日に続けて）行った後、コンピュータ画面に表示されたプロット図を保存した。図の下部の黒いウィンドウは、実行１および実行２のプロット図である。各場合における赤い線は全視野中のすべての細胞反応の平均を表す。ＭＣＦ−７ＷＴＣ（ＤＳ、実行１）およびＭＣＦ−７（ＭＤＲ、実行２）について記録されたこれら２つのプロット図中で示されているように、ＤＳ細胞（上方のプロット）のカルセイン蓄積量（２４０±８２、Ｎ＝３３６）はＭＤＲ細胞の蓄積量（１０３±８０、Ｎ＝１６１）より多かった。このことは、多数の細胞での平均を表すマルチウェルプレート実験で同じＤＳ細胞およびＭＤＲ細胞について観察されるカルセイン蓄積量の相対的差異と一致する。
【図１２】実施例４の実行１における各ＤＳ細胞の個々の細胞反応を示したグラフである。全細胞の反応の平均±標準偏差を黒色で重ねて示す。個々の反応は本グラフ中の個々のラインで示す。ＶＳＯＭ実験中、各細胞の細胞蛍光の平均値をモニターし、母集団平均からの偏差を検出した。例えば、矢印は、生存率および原形質膜完全性を失い始めた複数の細胞を示している。このような細胞はカルセインを保持することができず、蛍光信号が減少する。
【図１３】実行２（実施例４）で選択した４つの細胞について個々の細胞反応の測定結果を示す。これは、フローサイトメトリーおよびマルチウェルプレートアッセイでは得られない種類の詳細な生理学的情報の典型であり、本発明のＶＳＯＭ技術が望ましい能力を有していることを示す。示されているように、カルセインの蓄積および保持に関する詳細な比率情報は、ＶＳＯＭ実験全体にわたって入手可能である。
【図１４】本発明の１つの態様の図である。パネルＡはＤｅｅｐＶｉｅｗ Ｂｉｏｉｎｆｏｒｍａｔｉｃｓシステムの機能アーキテクチャの図である。パネルＢはポンプログ（ｐｕｍｐ
ｌｏｇ）の例を示し、パネルＣはＸＭＬレシピファイルの例を提供し、パネルＤはＩＣＳイメージヘッダーを提供する。
【図１５】インビボ実験およびインビトロ実験のフローチャートである。
【図１６】細胞の各区画について計算した構造および機能の特徴（ｆｅａｔｕｒｅ）ベースの階層保存を示す略図である。各実験は、微速度ビデオ撮影および、必要とされる灌流ポンプの操作とで構成される（図には示していない）。画像は最下層レベルに保存される。次のレベルには、細胞内の解剖学的組織の構造および機能が属性グラフとして保存される。第三のレベルでは、特定の位置に対応する反応曲線および、化合物の予想軌跡とが構成される。第四の層は、実験間の相互相関試験とともに推移確率を構築するための機構を提供する。
【図１７】培養皿中の細胞局在性の基準となるフレームを確立するための、エッチングを施した培養皿の使用を示した、拡大（ｃｌｏｓｅ−ｕｐ）略図である。[0001]
Portions of this invention have received government support from the U.S. Department of Energy (contract number DE-AC03-76SF00098). The United States Government has certain rights in the invention.
[0002]
This application claims the benefit of US Provisional Application No. 60 / 210,543, filed June 8, 2000, and US Provisional Application No. 60 / 290,755, filed May 14, 2001. Claims.
[0003]
Field of the invention
The present invention provides a method and apparatus for discovering and optimizing differences between cell types in a knowledge base. In particular, the present invention provides a visual servo controlled optical microscope and analysis method. The invention further provides a means for optimizing culture conditions for a particular cell subpopulation.
[0004]
Background of the Invention
Currently, in telepresence research, it is dominant to bring together specialists and facilities located in geographically dispersed locations (Hadida-Hassan et al., J. Struct. Biol., 125). : 229-234 [1999]; Parvin et al., "Visual Servoing for Online Facilities.
On-Line Facilities) ", IEEE Computer Mag. 56-62 [1997]; Potter et al., UltraMicros. , 77: 153-161 [1999]; Young et al. Supercomputer Appl. High Perf. Comput, pp. 170-181 [1996]; and Parvin et al., "A Collaborative Framework for Distributed Microscopy (A
Collaborative Framework for Distributed Microscopy) ", IEEE Conf. on
SuperComputing [1998]). This system is widely used in the field with collaborative frameworks (Parvin et al., [1997], supra). The focus of telepresence research has been on the remote functionality of the equipment and the automation required for large-scale data collection and analysis. MASH project at the University of California, Berkeley (McCanne, IEEE)
Internet Comput. 3: 33-44 [1999]) have developed a scalable multimedia architecture for collaborative applications in a fully distributed system using the MBone tool in a heterogeneous environment. The NCSA Havenero project achieves smooth management of shared information and simultaneous distribution to all clients in a computer-based centralized system described mainly in Java. Rutger University's DISCIPLE uses the CORBA framework for distributed access in a service-based centralized system for realizing a shared virtual space. Sun Microsystems (Sun)
Microsystem's Java Shared Development Tool Kit provides collaborative-aware Java code for transmitting data to participants in a communication session. This code supports three transport protocols: TCP / IP sockets, lightweight reliable multicast, and remote calling. In this framework, all objects are manageable and coordination takes place within a session that includes channels, tokens, blobs, and listeners. University of Michigan Upper
Atmosphere Research Collalaboratory (UARC) is a web-based distributed system, mostly written in Java. The system collects data from over 40 observation platforms for space physics research and is capable of both synchronous and asynchronous coordination. In this system, data providers publish their data on a data distribution server. Next, the client subscribes to the system to receive the desired information.
[0005]
Telepresence methods are used in biological sciences in addition to physical sciences. For example, in the age of post-genome sequencing, quantitative imaging of complex biological materials has become a significant challenge. In fact, the sequential measurements obtained with the various microscopy methods do not include a detailed analysis of multidimensional reactions (eg, reactions in time and space). Quantifying the simultaneous spatial and temporal behavior of multiple markers in a large population of multicellular aggregates is a method that is labor intensive, the absence of quantification tools, and the indexing of information It is difficult because they cannot do it. Ideally, a method that can track the kinetics and quantity, intracellular status, and morphological characteristics of a plurality of target proteins three-dimensionally using a large population is desirable.
[0006]
For example, thousands of antibodies and other reagents are available that can identify specific protein components in cells. Some antibodies can further distinguish functional variants of the protein that result from modifications such as phosphorylation status, protein conformation, and complex formation. Many intracellular proteins are involved in signal transduction pathways. Due to the complexity of potential events, the possibility of multiple modifications affecting protein function, and the lack of information about where and when the protein is active in signaling, these The details of the signal transmission path are not clear at present. Intrinsic biological variability and genomic instability are also factors that support the need for analyzes on large populations. Thus, there is a need for microscopy and image analysis methods that are useful for drawing more detailed images of cell signaling. This is especially true in the development of methods for diagnosing and treating diseases.
[0007]
Today, various diseases are understood at the molecular and genetic levels. Analysis of disease-related molecules is important in disease diagnosis and prognosis. However, the study of diseases such as cancer is currently limited by technology and model systems. Tissue samples typically contain multiple cell types (eg, abnormal cells, epithelial cells, stromal cells, endothelial cells, inflammatory cells, etc.), and this diverse population of cells is qualitative for protein and / or nucleic acid expression. Hinders quantitative or quantitative research. Interpret the significance of substantial denaturation of proteins or nucleic acids in typical specimens, since cells of interest (eg, tumor cells) often make up a relatively small percentage of the total cell population It is difficult. In addition, studies of cultured cells do not reveal details of the complex interactions that occur between cells. Furthermore, currently widely used techniques rely on methods such as tissue fixation, antigen-antibody recognition, and / or tissue staining, which typically result in the death of the cells being analyzed during sample processing. Inevitable. This limits the amount of information obtained about the cells in the specimen.
[0008]
In the study of cell physiology, it is necessary to set theoretical hypotheses about the phenomena observed in experiments, and the use of tools such as fluorescent probes and confocal microscopy to allow the intracellular disposition of various molecular species and intracellular structures. Can be decomposed in three dimensions. Often these hypotheses are then built as mathematical models. These models are then incorporated into simulations to correlate experimental results with phenomena observed in vivo.
[0009]
However, existing technologies have drawbacks. In general, these disadvantages stem from the current limitations of cell representation. In these methods, cells are represented as ideal and simple geometric shapes with spatially uniform behavior (ie, physiological features) and structure (ie, anatomical features). This hinders researchers trying to actually represent the observed physiological phenomena in simulations that easily correlate with actual experiments using intact cells. Therefore, validation of models and hypotheses is often very difficult.
[0010]
At present, much of the effort in modeling and simulating cell physiology has been directed to either highly specific models of individual mechanisms or abstract representations of more complex phenomena. Specific models include models of individual molecule interactions (eg, with respect to ion channels). Abstract models generally apply a simplification of the underlying mechanisms, which are only suitable for explaining a limited class of physiological problems and / or observations Is common. Thus, there is a need for methods and apparatus that facilitate real-time observation, correlating observed phenomena with disease states, and means for observing cells over time in situ.
[0011]
In addition, for some types of cancer (eg, certain leukemias and testicular cancers), chemotherapy has been successful and has given the patient a cure. However, very little progress has been made in improving the treatment of solid cancers (eg, breast cancer). Congenital or acquired multidrug resistance (MDR) seen in solid tumors is one of the key obstacles preventing cure with chemotherapy. Although in vitro chemosensitivity assays for evaluating drug response and predicting patient response have been developed for over 40 years, truly successful in vitro chemosensitivity tests have not yet been established. Thus, there remains a need for a reliable and meaningful in vitro chemosensitivity test.
[0012]
Summary of the Invention
The present invention provides a method and apparatus for discovering and optimizing differences between cell types in a knowledge base. In particular, the present invention provides a visual servo controlled optical microscope and analysis method. The present invention provides a means to closely monitor hundreds of individual living cells over time; quantification of dynamic physiological responses in multiple channels; real-time digital image segmentation and analysis; intelligent and repetitive, computer-applied cell stress. And cell stimulation; and provide the ability to return to the same cell field in long-term research and observation. The invention further provides a means for optimizing culture conditions for a particular cell subpopulation.
[0013]
In particular, the present invention relates to a method for automatically detecting and segmenting the field of view of cells contained in a cell array, and to computer control and external exposure of the response of cells to various stimuli, especially for population studies. Provides a way to calculate as a function of External exposure is controlled by the flow of compound in a computer-controlled syringe (or other suitable container) located near an optical device (eg, a microscope).
[0014]
In a preferred embodiment, VSOM provides a method for structurally delineating specific subcompartments of each cell in each selected field of view. The field of view is positioned by the user via the xy stage. In addition, VSOM is a specific subcellular compartment of each cell in the field of view (subcellular).
provides a means of measuring the response at the time of the experiment. In some embodiments, subcellular level responses are measured a) at the programmed stimulation frequency and b) as a function of cell exposure to compounds injected into the cell array. The present invention further provides control of a servo loop to adjust the exposure of the cell of interest as a function of the response measurement. The present invention offers maximum flexibility. For example, in some embodiments, external exposure is blocked when the average response of cells and / or intracellular compartments for all cells of interest in the field of view reaches a user-defined threshold. In another embodiment, external exposure is blocked when a desired percentage of cells and / or intracellular compartments reach a defined threshold. In yet another embodiment, external exposure is blocked when the average of the responses of a subset of cells and / or a subset of intracellular compartments reaches a particular profile. Each of these aspects can be repeated a desired number of times in any combination.
[0015]
Prior to the development of the present invention, it was technically difficult to obtain digital images of a large number of individual living cells under a microscope. It has also been difficult to simultaneously monitor multiple physiological responses within a cell over a long period of time without damaging the cell for many individual living cells. The first limitation was mainly due to the limited field of view at any magnification. The limited size and resolution of the solid state digital imager (ie, charge coupled device; CCD) inside the digital camera used also further limited the field of view. In the present invention, software control of the X, Y, Z microscope stage facilitates rapid, automatic, and repetitive monitoring of multiple fields of view (ie, large numbers of cells), thereby overcoming the aforementioned limitations. I do. The invention also facilitates simultaneous remote control of multiple microscopes by a central computer. In fact, memory, software associated with the computer (ie client) currently used to control local microscope peripherals (eg, scanning stage, filter wheel, perfusion pump, robotic arm, shutter, camera, etc.) , And processing resource limitations are overcome by the present invention. The present invention allows VSOM observations to be performed with remote control using the Internet and a more powerful central computer (server) located geographically away from remote users. The server performs or distributes more process intensive tasks and serves as a central location for a very large database of current and past cellular responses. Thus, the VSOM of the present invention recorded the benefits of network access to a powerful central server, as well as past observations on single cell responses and past correlations on multiple information channels. It takes advantage of the benefits of a large central database. In a particularly preferred embodiment, the VSOM system repeatedly detects cells, records and analyzes all cell responses observed for all channels, compares current cell responses, and correlates for each current information channel. In addition, an inquiry is made to a database in which past observation results on cell reactions and past observation results on correlations between information channels are recorded. In a single course of the experiment, the system then makes a series of decisions in the knowledge base on how to adjust the experimental parameters to achieve the goal of the experiment being performed.
[0016]
Thus, the present invention allows multiple complete experimental cycles to be performed during a single, relatively short course of experiment. This greatly accelerates the process of searching, finding, and optimizing differences between different cell types. The number of cells required is relatively small. In addition, since the response of individual cells is monitored, the sample for analysis may contain different types of cells. "Cell type" is achieved by any observation method that can divide a group of cells into multiple groups such that each cell in a group has similar characteristics that can be distinguished from cells in another group Classification. This classification is based on the results of repeated observations of relatively rapid physiological responses at the individual single-cell level (eg, observations of intracellular components, morphological changes, cell division, cell death, detachment, etc.). This is generally done. However, the term also includes various cell types that have been classified based on conventional histological criteria (eg, epithelial cells, tumor cells, muscle cells, adipocytes, nerve cells, etc.).
[0017]
Prior to the development of the present invention, automatic and knowledge-based decisions on how to change experimental parameters during the course of an experiment could not be made quickly when targeting large numbers of cells. The present invention provides the ability to perform a number of "test / store / analyze / learn / redesign / retest" cycles during the course of an experiment, and this capability has made many previously infeasible. Experiments and applications are possible. If the cells remain viable and the experiment remains in progress and online, the correlation between past and future responses to stimuli is maintained on a cell-by-cell basis. A relatively small amount of cell sample can be used to apply and repeat a variety and repetition of computer-generated stimuli during a single experiment. This is a great advantage when searching for and optimizing differences between cell types. The term "stimulus" refers to the application of a single stimulus, multiple stimuli, or a combination of multiple stimuli, once, repeatedly, or sequentially. Stimuli that trigger a physiological response include mechanical, physical, chemical, and biological entities. Accordingly, the present invention is not limited to any stimulus or stimulus type.
[0018]
Living cells can move or change morphology. Therefore, it is important to provide a means for analyzing morphometric features, texture, and other features of cells. Thus, upon returning to a field of view, the software needs to adjust the focus (z-axis control) and make small horizontal and vertical adjustments (x, y-axis control) to reacquire and re-register the field of view. There is. In addition, in order to track changes in cell position and shape, it is necessary to modify the original contours that define the outline of individual cells. Thus, in one aspect, the VSOM of the present invention provides a means for making decisions regarding device control based on the analysis of image content. In fact, the present invention provides the following means: means of computer automation to repeatedly and / or continuously apply stimuli to living cells; using microscopy and digital imaging, one subcellular or subcellular level to stimuli Means to repeatedly and / or continuously detect and / or monitor one or more physiological responses at the single cell level; suitable for reliable cell type identification and / or to optimize differences between cell types Means for rapidly observing the response of a sufficient number of individual cells to establish a suitable criterion for modifying the stimulus of the stimulus; allows knowledge-based stimulus control to be determined while the cells are available for observation and / or stimulation A means to rapidly analyze current and previously stored cellular responses (eg, database information); Means for monitoring over Yaneru.
[0019]
The invention further provides a means for defining the contours of living cells using transmitted light in addition to fluorescence. This facilitates tracking of cells that do not contain a fluorescent component. Furthermore, this makes it possible to reduce the amount of the fluorescently labeled compound that can exhibit cytotoxicity, and to avoid the strong fluorescence excitation irradiation necessary to generate fluorescence. Therefore, it is possible to observe cells for a longer time. Furthermore, cells that exhibit specific behavior can be collected individually at the end of the experiment and subjected to pooling (storage), establishment of cell lines, cloning, induction into specific differentiation states, analysis, and the like.
[0020]
In the device of the present invention, by rapidly cycling the imaging between transmitted light imaging and fluorescence imaging of various wavelengths, different physiological responses can be observed in different channels at substantially the same time. In addition to monitoring large numbers of cells, it is possible to observe multiple physiological responses during a single experiment.
[0021]
Using a multiwell fluorescent plate reader, it is possible to detect the ongoing fluorescent signal emitted by the entire population of living cells, but no single cell response can be observed. However, in one aspect of the VSOM of the present invention, the discovery and optimization of a fluorescent assay suitable for a multiwell plate reader is possible. In addition, according to a general technique, a multi-channel fluorescent signal emitted from an individual living cell can be instantaneously detected using flow cytometry. Cells can be sorted and recovered according to the fluorescent signal. However, the cells pass through the detector at a high speed, and the resulting measurement is single (or possibly multi-channel). Individual cells are not tracked over a fixed amount of time, and repeated observations of the same cell, correlating past and current responses, are not possible. The present invention overcomes these limitations and facilitates the development and optimization of fluorescent assays suitable for flow cytometry and active cell sorting.
[0022]
The invention has application in cell type-specific fluorescence assays useful for any type of cell (eg, animal cells, plant cells, microbial cells, etc.). Thus, the invention is not limited to any type of cell or to any particular fluorescence or other assay system. In a preferred embodiment, human cells are targeted, but in other embodiments, cells obtained from other mammals and plants and microorganisms are also targeted. For example, the invention provides a means for detecting and identifying normal cells, pre-malignant cells, malignant cells, and / or multidrug resistant cancer cells obtained from a tissue. Further, the invention provides a means for designing a protocol for selectively labeling rare cell types (eg, stem cells or fetal cells in maternal blood). In some embodiments, these protocols are modified for use with radioactive and other labeling systems. The present invention also facilitates the design and application of assays to establish chemotherapeutic regimens (including drug combinations) tailored to individual patients and / or individual tumors of the patients. The invention further provides assays useful in searching for a large number of potential drugs, pesticides, herbicides, and other compounds for various uses in medicine, agriculture, biotechnology, and the like. Further, the invention provides assays useful in searching for a large number of candidate agents for use in cell proliferation, cytotoxicity, and / or differentiation.
[0023]
The present invention further provides specific growth by favoring the cell type of interest or by designing environmental conditions or protocols that act cytotoxic to other cell types present in culture. It provides a means of designing in vitro tissue culture conditions that result in proliferation of a cell type. Thus, the present invention provides a means for culturing cells that are often difficult to grow in vitro. Further, the present invention provides a means for creating cell culture conditions suitable for inducing cells to a specific differentiation endpoint. For example, the present invention can be used to develop the conditions necessary to induce stem or embryonic cells to a desired end point.
[0024]
The invention also has application in the development and performance of various rapid in vitro tests. For example, if there are multiple candidate tissue or bone marrow donors for a patient, cells from different donors are mixed with the patient's immune cells and immune rejection is observed at the single cell level. Thus, the identification of the optimal donor for that particular patient is facilitated. In addition, these tests can be performed in the presence of drugs (or drug candidates) designed to suppress transplant rejection, and in conjunction with optimal donor selection for patients, administration of optimal anti-rejection drugs. The choice of regimen (ie, single agent or multiple agent) is facilitated.
[0025]
The present invention includes receiving cell image data (ie, information regarding a cell's physiological, morphological, and / or other characteristics) from a detection device that monitors a cell or an intracellular component of the cell; Analyzing the data; and automatically activating a plurality of stimulators adapted to stimulate a cell or an intracellular component in response to the analyzed cell image data. In some embodiments, the invention further comprises automatically and independently operating a plurality of stimulators to stimulate cells in response to the analyzed cell image data. In yet another aspect, analyzing the cell image data includes at least one of cell size analysis, cell color analysis, and cell movement analysis. In some preferred embodiments, the detection device comprises a visual servo-controlled optical microscope (VSOM), in which case the method further comprises monitoring the cells using the VSOM. In another aspect, the stimulator is a syringe. In yet another aspect, the method further comprises storing the cell image data as a function of time. In some particularly preferred embodiments, the cell is a living cell and the intracellular component is a component within a living cell. In another aspect, the method further comprises repeating the foregoing steps a desired number of times and in a desired order.
[0026]
The present invention also provides a cell array; detecting light emitted by the cell array and receiving a first signal corresponding to the light generated by the cell array, receiving the first signal and using the first signal to stimulate the cell array; An optical system for providing a computer with a second signal for controlling delivery of at least one test stimulus to the cell array to obtain a cell array; and receiving the second signal and delivering the test stimulus A method and system for analyzing a cell array, including a robotic system, wherein the optical system further detects light emitted by the unstimulated cell array. In some embodiments, the test stimulus is automatically provided to the cell array according to the first signal. In some preferred embodiments, the robotic system delivers the test stimulus to the cell array via at least two microfluidic channels. In yet another aspect, the optical system includes at least one microscope. In some preferred embodiments, the microscope is a confocal microscope, and in another embodiment, the microscope is a fluorescence microscope. In some embodiments, the cell array comprises at least two cell types. In yet another aspect, the optical system scans the cell array to detect intracellular components in cell types included in the cell array. In some preferred embodiments, the intracellular component is selected from the group consisting of a nucleus, a nucleolus, a mitochondria, a lysosome, a phagolysosome, and a storage vesicle. In some particularly preferred embodiments, at least one intracellular component is labeled. In another preferred embodiment, the intracellular component is labeled with a test stimulus. In some embodiments, the intracellular component is labeled with at least one compound selected from the group consisting of a fluorescent dye, a radioactive compound, an enzyme, and a vital dye. In another aspect, the optical system converts the sign into digital data. In yet another aspect, the robotic system utilizes digital data to automatically coordinate the delivery of a bioactive component to a cell array. In another aspect, the invention provides a means for remote control of a computer system. In yet another aspect, the invention provides a means for comparing cell information (eg, information in a result profile) to a reference database containing previously acquired cell information and data.
[0027]
The present invention also provides a code for receiving cell image data from a detection device that monitors cells or intracellular components; a code for analyzing cell image data; and a cell or cell in response to the analyzed cell image data. A computer readable medium is provided that includes code for automatically adjusting a plurality of stimulators adapted to stimulate internal components. In some preferred embodiments, the computer readable medium further includes code for controlling a visual servo controlled light microscope (VSOM) that monitors the cells. In yet another aspect, the computer-readable medium further comprises at least one of a code for analyzing cell size, a code for analyzing cell color, and a code for analyzing cell movement. Including. In yet another aspect, the computer-readable medium further includes code for automatically and independently operating a plurality of stimulators for stimulating cells in response to the analyzed cell image data. In yet another aspect, the stimulator is a syringe. In some preferred embodiments, the computer readable medium further includes code for storing the cell image data as a function of time. In yet another preferred embodiment, the cell is a living cell and the intracellular component is a component within a living cell.
[0028]
General description of the invention
The present invention combines the processing, memory, and analytical capabilities of most modern network computers with the complex and interactive biological and biochemical pathways or "circuitry" of a single living cell. Provide the means for: Further, this process can be repeated for hundreds or thousands of cells, and humans do not engage in this connected and automatic system. A communication or feedback loop is established when living cells, interrogated by the stimuli and perturbations of the computer, "react" in a manner that can be detected and analyzed by the computer. These complex stimuli and perturbations due to computer application can be applied repeatedly and can be altered appropriately based on the complex "response" received from the cell.
[0029]
Some embodiments of the present invention rapidly discover subtle differences in the net physiological response of different cell "species" and / or between cells of the same "species" in different physiological states. It has the use of doing. The present invention provides a series of additional tests or stimuli designed to segregate or localize the observed differences to specific biochemical pathways, expression of specific genes or gene sets, or other specific differences. This provides a means for quickly, automatically and efficiently refining or optimizing the difference in net response of different cell types. In this manner, subtle differences in biological responses between two or more different populations of living complex systems (live cells), because certain differences are identified and targeted in a very precise manner. Can be amplified.
[0030]
The present invention provides a method and apparatus for discovering and optimizing differences between cell types in a knowledge base. In particular, the present invention provides a visual servo-controlled light microscope (VSOM) and analysis method for the automatic generation and subsequent automatic analysis of large, defined arrays of living cells. The invention further provides a means for searching, quantifying, and optimizing differences in the response of cells to an applied stimulus or combination of stimuli. The present invention can take advantage of known differences in cell responses and / or search for and discover new differences in cell responses in an automated manner to achieve various objectives. Thus, according to the present invention, it is possible to discover and apply a stimulus or a combination of stimuli to obtain the desired difference in cellular response.
[0031]
In addition, the present invention allows for the identification of individual cell types and the physiological state of individual cells for a large number of individual cells attached or firmly adhered to a surface. Accompanying this identification of individual cells is the location of the cells relative to a number of neighboring cells. In this manner, a large array of living cells arbitrarily placed on a surface is automatically converted to a large array of identified cells at defined x, y, z locations. Such arrays are then monitored and tracked during subsequent VSOM stimulation, perturbation, and interrogation experiments. VSOM stimulation, perturbation, and interrogation include the use of fluorescent probes, the addition of chemical and biological agents, and the application of physical stimuli. These, applied alone or in combination in various amounts, intervals and continuity, are automatically recorded, analyzed and optimized. Such stimuli, perturbations, and interrogations can be benign, can cause lethal stress on the cell, or have some intermediate effect on the cell. You may. In accordance with the present invention, the biological consequences of the physiological response to automatically applied stimuli, perturbations, and interrogations during one experimental course can be determined on a cell-by-cell and organelle-by-cell organelle basis. Can be quantitated, monitored, and stored over time, and correlated with other reactions.
[0032]
The history of applied stimuli, perturbations, and interrogations, as well as the correlated multiple individual cell histories of hundreds or thousands of cells observed simultaneously on multiple channels, can be monitored remotely or locally. Stored in the database. In addition, the final biological consequences of such stimuli are collected for subsequent VSOM experiments. As this database grows, VSOM assays and protocols become more predictive and more useful in controlling and optimizing instruments and processes.
[0033]
In the present invention, stimulus, search, optimization, and VSOM instrument control are performed in real-time during the VSOM experimental process, based on knowledge of past cellular responses to applied probes, stimuli, perturbations, and interrogations. It is possible to access the database locally or remotely. The present invention allows for the development of various in vitro and in vivo assays specific to the cell type and / or physiological state of the cell. The development of assays and methods based on the methods and systems of the invention and the accumulation of stored knowledge allows for faster, more benign, and specific, more efficient use of automated generation of large-scale live cell arrays. Assay is possible. This allows for the rapid and automatic generation of minimally stimulated or perturbed live cell arrays, and then searches for new or improved recipes and protocols that take advantage of cell type differences. A valid cycle is provided that can be used for VSOM. Next, this process, instrument control knowledge, and the resulting recipes and protocols to achieve cell-specific physiological responses and cell-specific biological consequences for the purpose of achieving various objectives. The recipes and protocols that are generated can be used in VSOM and non-VSOM environments.
[0034]
These purposes include, but are not limited to, (i)-(xiii): (i) rapid and automated in vitro culture conditions and protocols that favor only a specific subpopulation of cells. (Ii) developing cell type specific assays, protocols, and diagnostic kits suitable for non-VSOM instruments or situations; (iii) cell type specific for VSOM instruments only Developing the assay, (iv) discovering specific differences between cell types that are targeted for drug development, (v) obtaining the desired differentiation, such as obtaining various specific cell types from stem cells. Discovering in vitro conditions and protocols for obtaining cells in a state, (vi) specific genes or gene sets in a complex and interactive biochemical network of living cells (Vii) discovering the relationship between the expression or non-expression of a gene or set of genes and the resulting cell phenotype, (viii) the complex and interactive nature of individual living cells Discovering the effects of drugs or other substances on complex biochemical networks, (ix) finding substances and protocols suitable for cell-specific delivery of drugs or probes, (x) tumors of individual patients, especially Finding chemotherapeutic regimens designed for tumors that may exhibit multidrug resistance, (xi) finding the relative chemosensitivity or toxicity of a substance to a particular individual, (xii) identifying Discovering the suitability of a particular organ donor for an individual, and (xiii) 3D imaging of the expression of a particular gene in a living human And such as those utilizing radiolabeled probe, to develop an in vivo assay. The present invention further provides a means for specifically identifying a particular cell present in a culture multiple times, even when an incubation period is provided between each analysis (ie, the culture is placed in a cell culture incubator). Provides a means of repeatedly analyzing cells over time.
[0035]
In one embodiment of the VSOM approach, a solution containing individual dispersed live cells (or very small clumps of cells or tissue pieces) is placed in a sterile container (ie, in vitro). Cells will settle to the bottom of the container unless there is a means to attach the cells to the container wall, and then (if conditions are correct) begin to attach to the bottom of the container. During this process, the cells flatten and then (if conditions are right) begin to grow, divide, and proliferate. In some cases, the bottom of the container needs to be treated so that, for example, round cells that do not show any tendency for flattening and surface attachment will adhere to the bottom of the container and will not be washed away by liquid exchange.
[0036]
In some preferred embodiments, the present invention uses a microscope, typically an inverted microscope, with a magnifying lens (ie, objective lens) located below the container containing the cell solution. A digital camera is attached to the microscope to automatically acquire digital images, and (in some cases) cells are illuminated with standard white light from a standard halogen bulb. Next, cell images suitable for digital camera imaging are obtained using standard transmitted light techniques (phase contrast method, differential interferometry, bright field method, dark field method, etc.). In some cases, additional images are acquired continuously and at high speed using very bright illumination, such as from a mercury or xenon arc lamp. This light passes through one or more optical filters contained in a computer-controlled rotating optical filter wheel, which irradiates the cells with filtered violet, blue, or yellow light and produces a series of digital images. Acquired continuously. In a microscope appropriately equipped for surface fluorescence microscopy, cells containing one or more fluorescent probes emit a specific color of fluorescence when irradiated with the appropriate color of fluorescent excitation light. Thus, a continuous digital image obtained by changing the position of the optical filter wheel results in a separate continuous digital image or digital "channel" of the blue, green and red fluorescence emitted by the cells. In this manner, the amount and relative distribution of (for example) blue, green, or red fluorescent probes in cells is monitored with digital images of blue, green, or red fluorescent emission (blue, green, and red channels). The overall shape and texture of the cells can be observed in transmitted light digital images (transmitted light channels).
[0037]
The relative intensity and color of the fluorescence observed in different subcellular compartments can indicate the rate of uptake and retention of different fluorescent probes, and how this fluorescence intensity is disrupted and redistributed within the cell Provides important physiological, anatomical, and morphological information about Morphological information is also available in transmitted light digital images. Changes that occur in any or all of these parameters as a function of time or the applied stimulus are the response of the cell to the fluorescent probe used and to other stimuli or perturbations applied to the cell or its environment. ] Can be interpreted. In addition, separate reactions are separated into multiples so that one reaction can be observed in the red channel, one reaction in the green channel, one reaction in the blue channel, and one reaction in the transmitted light channel. Can be monitored and tracked on digital channels. In this manner, physiological responses to computer applied stimuli, perturbations, and interrogations are encoded in the continuous digital image. These responses are decoded by an algorithm that extracts the information contained in each digital image as it passes through the computer and detects the response of the cells, in a sense, the "response" to the computer applied stimulus.
[0038]
In one particularly preferred embodiment of the present invention, the response or "response" of the cell to a computer applied interrogation (or "interrogation") in the form of stress, stimulation, inhibition or perturbation is determined during the course of the experiment. It is decoded and interpreted quickly (ie, "in real time"). Therefore, additional "questions" can be asked based on previous "responses" or interpretations of cellular responses. This process forms a communication loop between the cell and the computer, and the current biological state of individual living cells and the current biological “circuitry” (eg, current biological and biochemical This corresponds to a means for “reading out” information on a typical process.
[0039]
One means of applying computer controlled stimulation, perturbation, and interrogation is through a computer controlled syringe pump. In this manner, various fluorescent probes, as well as solutions of chemicals and biologicals, can be passed through the attached live cells, and when the substances and probes are added to or from the cell specimen container. As the cells are washed away, the response of the cells to these substances can be monitored. Such a process may be described as a change or perturbation in the cellular microenvironment. There are other physical means to perturb the cellular environment, such as changing the temperature of the cell container, irradiating the cells with strong UV light, or applying an electric field to the cells in the container. Such environmental variations may be controlled by a computer and do not require a syringe pump. In some particularly preferred embodiments, the user is provided with a computer capable of tracking a vast number of past combinations of such stimuli, as well as how they were applied and how the cells responded. Make use of your abilities. This may be done not only by monitoring ongoing experiments, but also by referring to a very large online computer database that records past cell stimuli and responses. The ability to access such large databases during the course of real-time experiments with live cells is a significant advantage over widely used methods.
[0040]
Furthermore, means for identifying, identifying and classifying living cells of different species are provided by specially modulated (generated in a computer-controlled manner) cell type-specific and time-varying dynamic physiological responses. The present invention provides a means to discover and refine such unique cell type-specific "physiological fingerprints" to the point where living cells are minimally invasive and relatively benign. . Using morphometric parameters (obtained in the transmitted light channel in the absence of a fluorescent probe) to quantify the response to benign perturbations of the environment is an example of a highly refined identification assay.
[0041]
One of the challenges that was overcome during the development of the present invention was that standard digital image analysis techniques for delineating living cells and their individual subcellular compartments were often used to study living cells. The problem is that it is insufficient. These algorithms often require that the cells be physically separated (ie, that they do not touch or overlap) and that they are very evenly labeled with relatively bright fluorescent dyes. Is also required. These algorithms fail if the staining is non-uniform (i.e., there are cells that are brighter than other cells) or if the illumination pattern is non-uniform. In many cases, the prolonged fluorescent exposure required for the success of these algorithms will act toxic to living cells.
[0042]
Upon successful detection and profiling of individual cells (and ideally, a particular subcellular compartment) in a digital image by a suitable computer algorithm, the change in the amount of fluorophore that occurs in response to a computer applied stimulus (ie, Fluorescence microscopy) or morphological changes can be quantified, recorded and analyzed by computer. The computer can then make decisions based on the results of these analyses, with the aim of extracting additional information from the living cells, and instruct the syringe pump or other computer controlled microscope peripherals. Can be sent. This additional information and all subsequent information obtained in subsequent cycles is useful in optimizing and amplifying the first small differences between cell types observed after a single stimulus.
[0043]
Because of the normal heterogeneity and variability between individual living cells (even of the same type), it is difficult or impossible to interpret the response of living cells unless a large number of cells are observed at the same time. There are many. The need for observing large numbers of cells and concomitantly the need to rapidly process large amounts of digital data is also solved by the present invention. Local computers usually do not have such computing power, and if they do, the cost of the equipment will increase significantly. Thus, some aspects of the invention provide for controlling a device by a more powerful remote computer. In addition, the required number of cells may not always be visible within a single digital image or single microscope field. Thus, another aspect of the present invention provides the ability to observe more cells by acquiring additional digital images in close proximity.
[0044]
The present invention has also made it possible to observe a large number of cells in the following manner. The biological response and consequences to the applied stimulus are not always rapid, but are often relatively slow. This is especially true if the stimulus is applied in a controlled, slow manner. Therefore, according to the present invention, it is possible to continuously bring different cell groups into the field of view of the digital camera. When monitoring five groups of cells, the computer-controlled microscope stage moves continuously and repeatedly between five different regions of the specimen container, always returning to the same five cell population at regular intervals throughout the experiment. This precision is not provided by the precise xy stage alone, but rather involves detecting cell contours and reacquiring the same random distribution pattern representing a particular population of attached cells, and It is also provided by the ability of the computer algorithm to make fine adjustments to the xy stage as needed until it is properly positioned again for digital imaging.
[0045]
Some examples of the above concept are schematically illustrated in various figures (see, for example, FIGS. 4 and 5). Examples of necessary equipment and protocols are also detailed herein.
[0046]
Many aspects of the present invention can be demonstrated by multi-view (MF) VSOM experiments, as shown schematically in FIG. Four computer controlled syringe pumps (A, B, C, D) inject various solutions into a circular microenvironment chamber containing live cells and one or more environmental sensors. The solution enters the cell chamber where two physically separated patches of cells are located. In-line flow sensors I and II monitor the composition of the injected solution and sensors III and IV monitor environmental conditions in the chamber. In this diagram, the cells on the left represent normal cells (hatched circles) and cancer cells (open circles) as expected (in this simplified model) for breast cancer biopsy specimens. Represents a heterogeneous mixture containing The cells on the right (hatched circle) represent a biopsy specimen obtained from normal tissue (eg, contralateral breast). A VSOM controlled xy stage repeatedly repositions the cell chamber above the microscope objective (located below the chamber) so that the same set of cells (1, 2, 3, and 4) are repeatedly imaged. The response of these living cells is quantified continuously as various perturbations are applied. These perturbations include infusion of solutions of various compositions and control of other perturbations that cannot be administered as a solution. These perturbations may include gas, temperature fluctuations, irradiation, electric or magnetic fields, or other physical perturbations applied with the VSOM algorithm or recipe. In this figure, the outflow from the environmental chamber is also monitored with in-line flow sensors V, VI, VII, and VIII. Cell responses are analyzed in real time using both fluorescent and transmitted light modes. To achieve specific experimental goals, the appropriate syringe and appropriate environmental conditions are controlled by the VSOM based on the VSOM analysis that is being performed on individual responses to perturbations and environmental changes. In the case shown in this figure, it is also possible to place a microcapillary near or in a specific cell.
[0047]
FIG. 4 is a schematic diagram illustrating one embodiment of a VSOM experiment. At the start of the experiment (time point 1), cell types A and B are randomly located on the surface and are present as unidentified cells in a heterogeneous array. At this first point, the cells are not identified and are called "a", "b" and "?". The VSOM system scans a series of solutions (from syringes i-v) while searching for a protocol that selectively labels cell type A with a fluorescent dye (this figure shows that this occurred at time point 5). Add and apply a series of physical stimuli. At time point 5 (indicated by a check mark in the figure), the two cell types and the success protocol are identified and the cell response is recorded in a database. In some cases, the experiment is terminated at time 5 when a new protocol for selectively labeling cell type A in the presence of cell type B is discovered. Next, a successful protocol is identified as each step performed between time 3 and time 5.
[0048]
In the subsequent experiments, only the steps between time 3 and time 6 were performed so that the identification of all cells of type A and type B and the recording of the position of each cell in the array were completed at time 6. In some cases, in an optimized manner, it is performed on the cells and the fluorescent dye is washed away. In this manner (possibly after further refinement of the labeling protocol), an array of identified cells with minimal perturbation and free of foreign compounds or markers is obtained. The next VSOM experiment may then be initiated using the VSOM generated array of identified living cells.
[0049]
The sequence of events shown in FIG. 4 panel A contains a large amount of experimental information and results, which can be described in more detail. For example, between time 1 and time 2, the cells undergo physical perturbation (such as a decrease in temperature; indicated by a white arrow above the upper time axis labeled "physical stimulus" in the figure). And then inject the solution in syringe (i). Regarding the solution injection, the pump (i) was operated for a certain period of time (indicated by the width of the hatched rectangle on the time axis) and a constant flow rate (shown by the hatched line) in the graph labeled “pump operation” in this figure. (Rectangular height). At time point 2, morphological changes occur, which have been continuously quantified for each cell using image analysis techniques. However, this cellular response is the same for both cell types. Therefore, at this point, no discrimination between cell types has been performed, and an automatic search is continuously performed. Between time points 2 and 3, the solution in syringe (iii) is injected, followed by physical stimulation (such as UV light irradiation). At time 3, both cell types still behave similarly and the experiment is continued. The solution is injected from syringe (v) between time points 3 and 4 so that both cell types are labeled with a fluorescent dye (shown in gray shade), followed by a physical stimulus (such as applying an electric field). The solution of syringes (iv and ii) is then added (almost) between time points 4 and 5. As a result, the cell type B loses the fluorescent dye earlier than the cell type A, so that at the time 5, only the cell type A is in a labeled state. Between point 5 and point 6, the solution is injected at a higher flow rate from syringe (ii) until none of the cells contain any fluorescent dye (point 6). In this manner, the entire array of living cells is identified and none of the cells contains foreign compounds or fluorescent probes. Subsequent identification assays can minimize the history of cell exposure and stimulation.
[0050]
The time-dependent responses of the cell types A and B detected by the computer by segmentation and analysis of the digital image may then be displayed as in FIG. 4 Panel B. In this figure, a schematic of the cells is represented by a transmitted light channel (the y-axis represents morphological measurements representing the shape or texture of the cell) and a monochromatic fluorescent channel (the y-axis represents the average of the cell fluorescence intensity). It has been replaced by a computer-detected and quantified reaction. The reaction of cell type A is shown by a solid line, and the reaction of cell type B is shown by a dotted line. FIG. 4 Panel B illustrates one important concept of the present invention, namely the fact that the computer has the ability to correlate each channel of reaction information. One such example is seen near time point 4 on panel B of FIG. That is, in this vicinity, for example, changes in the texture of the cell membrane may indicate a morphometric response that can be correlated with the loss of fluorescent dye from the cells. After this phenomenon has been confirmed, assays that do not require the presence of fluorescent dyes to detect cellular reactions. The cellular response can be identified by the morphometric response profile of the cell.
[0051]
Thus, the present invention provides a method and apparatus for discovering and optimizing differences between cell types in a knowledge base. In particular, the present invention provides a visual servo controlled optical microscope and analysis method as described above. The present invention provides a means to closely monitor hundreds of individual living cells over time; quantification of dynamic physiological responses in multiple channels; real-time digital image segmentation and analysis; intelligent and repetitive, computer-applied cell stress. And cell stimulation; and provide the ability to return to the same cell field in long-term research and observation. The invention further provides a means for optimizing culture conditions for a particular cell subpopulation.
[0052]
Prior to the development of the present invention, it was technically difficult to obtain digital images of a large number of individual living cells under a microscope. It has also been difficult to simultaneously monitor multiple physiological responses within a cell over a long period of time without damaging the cell for many individual living cells. The first limitation was mainly due to the limited field of view at any magnification. The limited size and resolution of the solid state digital imager (ie, charge coupled device; CCD) inside the digital camera used also further limited the field of view. In the present invention, software control of the x, y, z microscope stage facilitates rapid, automatic, and repetitive monitoring of multiple fields of view (ie, large numbers of cells), thereby overcoming the limitations described above. I do. The invention also facilitates simultaneous remote control of multiple microscopes by a central computer. In fact, memory, software associated with the computer (ie client) currently used to control local microscope peripherals (eg, scanning stage, filter wheel, perfusion pump, robotic arm, shutter, camera, etc.) , And processing resource limitations are overcome by the present invention. The present invention allows VSOM observations to be performed with remote control using the Internet and a more powerful central computer (server) located geographically away from remote users. The server performs or distributes more process intensive tasks and serves as a central location for a very large database of current and past cellular responses. Thus, the VSOM of the present invention recorded the benefits of network access to a powerful central server, as well as past observations on single cell responses and past correlations on multiple information channels. It takes advantage of the benefits of a large central database. In a particularly preferred embodiment, the VSOM system repeatedly detects cells, records and analyzes all cell responses observed for all channels, compares current cell responses, and correlates for each current information channel. In addition, an inquiry is made to a database in which past observation results on cell reactions and past observation results on correlations between information channels are recorded. In a single course of the experiment, the system then makes a series of decisions in the knowledge base on how to adjust the experimental parameters to achieve the goal of the experiment being performed.
[0053]
Thus, the present invention allows multiple complete experimental cycles to be performed during a single, relatively short course of experiment. This greatly accelerates the process of searching, finding, and optimizing differences between different cell types. The number of cells required is relatively small. In addition, since the response of individual cells is monitored, the sample for analysis may contain different types of cells. "Cell type" is achieved by any observation method that can divide a group of cells into multiple groups such that each cell in a group has similar characteristics that can be distinguished from cells in another group Classification. This classification is based on repeated observations of relatively rapid physiological responses at the level of individual cells (eg, observations of intracellular components, morphological changes, cell division, cell death, detachment, etc.). This is generally done. However, the term also includes various cell types that have been classified based on conventional histological criteria (eg, epithelial cells, tumor cells, muscle cells, adipocytes, nerve cells, etc.).
[0054]
Prior to the development of the present invention, automatic and knowledge-based decisions on how to change experimental parameters during the course of an experiment could not be made quickly when targeting large numbers of cells. The present invention provides the ability to perform a number of "test / store / analyze / learn / redesign / retest" cycles during the course of an experiment, and this capability has made many previously infeasible. Experiments and applications are possible. If the cells remain viable and the experiment remains in progress and online, the correlation between past and future responses to stimuli is maintained on a cell-by-cell basis. A relatively small amount of cell sample can be used to apply and repeat a variety and repetition of computer-generated stimuli during a single experiment. This is a great advantage when searching for and optimizing differences between cell types. The term "stimulus" refers to the application of a single stimulus, multiple stimuli, or a combination of multiple stimuli, once, repeatedly, or sequentially. Stimuli that trigger a physiological response include mechanical, physical, chemical, and biological entities. Accordingly, the present invention is not limited to any stimulus or stimulus type.
[0055]
Living cells can migrate or change morphology. Thus, upon returning to a field of view, the software needs to adjust the focus (z-axis control) and make small horizontal and vertical adjustments (x, y-axis control) to reacquire and re-register the field of view. There is. In addition, in order to track changes in cell position and shape, it is necessary to modify the original contours that define the outline of individual cells. Thus, in one aspect, the VSOM of the present invention provides a means for making decisions regarding device control based on the analysis of image content. In fact, the present invention provides the following means: means of computer automation to repeatedly and / or continuously apply stimuli to living cells; using microscopy and digital imaging, one or more subcellular or subcellular levels to stimuli Means to repeatedly and / or continuously detect and / or monitor one or more physiological responses at the single cell level; suitable for reliable cell type identification and / or to optimize differences between cell types Means for rapidly observing the response of a sufficient number of individual cells to establish a suitable criterion for modifying the stimulus of the stimulus; allows knowledge-based stimulus control to be determined while cells are available for observation and / or stimulation A means to rapidly analyze current and previously stored cellular responses (eg, database information); Means for monitoring over Yaneru.
[0056]
The invention further provides a means for defining the contours of living cells using transmitted light in addition to fluorescence. This facilitates tracking of cells that do not contain a fluorescent component. Furthermore, this makes it possible to reduce the amount of the fluorescently labeled compound that can exhibit cytotoxicity, and to avoid the strong fluorescence excitation irradiation necessary to generate fluorescence. Therefore, it is possible to observe cells for a longer time. Furthermore, cells that exhibit specific behavior can be collected individually at the end of the experiment and subjected to pooling (storage), establishment of cell lines, cloning, induction into specific differentiation states, analysis, and the like.
[0057]
In the device of the present invention, by rapidly cycling the imaging between transmitted light imaging and fluorescence imaging of various wavelengths, different physiological responses can be observed in different channels at substantially the same time. In addition to monitoring large numbers of cells, it is possible to observe multiple physiological responses during a single experiment.
[0058]
Using a multiwell fluorescent plate reader, it is possible to detect the ongoing fluorescent signal emitted by the entire population of living cells, but no single cell response can be observed. However, in one aspect of the VSOM of the present invention, the discovery and optimization of a fluorescent assay suitable for a multiwell plate reader is possible. In addition, according to a general technique, a multi-channel fluorescent signal emitted from an individual living cell can be instantaneously detected using flow cytometry. Cells can be sorted and recovered in response to the fluorescent signal. However, the cells pass through the detector at a high speed, and the resulting measurement is single (or possibly multi-channel). Individual cells are not tracked over a fixed amount of time, and repeated observations of the same cell, correlating past and current responses, are not possible. The present invention overcomes these limitations and facilitates the development and optimization of fluorescent assays suitable for flow cytometry and active cell sorting.
[0059]
The invention has application in cell type-specific fluorescence assays useful for any type of cell (eg, animal cells, plant cells, microbial cells, etc.). Thus, the invention is not limited to any type of cell or to any particular fluorescence or other assay system. In a preferred embodiment, human cells are targeted, but in other embodiments, cells obtained from other mammals and plants and microorganisms are also targeted. For example, the invention provides a means for detecting and identifying normal cells, pre-malignant cells, malignant cells, and / or multidrug resistant cancer cells obtained from a tissue. Further, the invention provides a means for designing a protocol for selectively labeling rare cell types (eg, stem cells or fetal cells in maternal blood). In some embodiments, these protocols are modified for use with radioactive and other labeling systems. The present invention also facilitates the design and application of assays to establish chemotherapeutic regimens (including drug combinations) tailored to individual patients and / or individual tumors of the patients. The invention further provides assays useful in searching for a large number of potential drugs, pesticides, herbicides, and other compounds for various uses in medicine, agriculture, biotechnology, and the like. Further, the invention provides assays useful in searching for a large number of candidate agents for use in cell proliferation, cytotoxicity, and / or differentiation.
[0060]
The present invention further provides specific growth by favoring the cell type of interest or by designing environmental conditions or protocols that act cytotoxic to other cell types present in culture. It provides a means of designing in vitro tissue culture conditions that result in proliferation of a cell type. Thus, the present invention provides a means for culturing cells that are often difficult to grow in vitro. Further, the present invention provides a means for creating cell culture conditions suitable for inducing cells to a specific differentiation endpoint. For example, the present invention can be used to develop the conditions necessary to induce stem or embryonic cells to a desired end point.
[0061]
The invention also has application in the development and performance of various rapid in vitro tests. For example, if there are multiple candidate tissue or bone marrow donors for a patient, cells from different donors are mixed with the patient's immune cells and immune rejection is observed at the single cell level. Thus, the identification of the optimal donor for that particular patient is facilitated. In addition, these tests can be performed in the presence of drugs (or drug candidates) designed to suppress transplant rejection, and in conjunction with optimal donor selection for patients, administration of optimal anti-rejection drugs. The choice of regimen (ie, single agent or multiple agent) is facilitated.
[0062]
DETAILED DESCRIPTION OF THE INVENTION
The present invention provides a method and apparatus for discovering and optimizing differences between cell types in a knowledge base. In particular, the present invention provides a visual servo controlled optical microscope (VSOM) and analysis method. The present invention provides a means to closely monitor hundreds of individual living cells over time; quantification of dynamic physiological responses in multiple channels; real-time digital image segmentation and analysis; intelligent and repetitive, computer-applied cell stress. And cell stimulation; and provide the ability to return to the same cell field in long-term research and observation.
[0063]
1 and 2 are schematic diagrams illustrating the three main components of the preferred embodiment of the present invention. These include (i) an optical platform for digitally imaging live cells in a cell chamber, (ii) a physical platform, often attached to or near the optical platform, referred to as a "robot component". It consists of a computer controlled peripheral and (iii) a local host computer ("client") used for local commands of the robot peripheral. In another aspect, the present invention further provides a remote computer (or "server") having software for remote control of VSOM equipment and operating over the Internet.
[0064]
A. Optical platform, robot components, and control electronics
In one aspect of the invention, the VSOM optical platform is an inverted optical microscope configured to perform both transmitted light illumination and surface fluorescent illumination on a specimen mounted in the center of the microscope stage. A plurality of objective lenses are mounted below the microscope stage for image generation and magnification. In one embodiment, four robot components (ie, computer controlled microscope peripherals) are also located near the microscope stage. In a particularly preferred embodiment, a digital camera with an internal camera shutter is mounted on one of the upper camera ports on the microscope. The microscope stage is a computerized XY scanning stage and includes the position of two computer controlled shutters where the top shutter controls transmitted light illumination and the internal rear shutter controls surface fluorescent illumination. In addition to the internal shutter, it has a computer-controlled six-position rotating optical filter that can illuminate and fluorescently excite the specimen with light of defined spectral components.
[0065]
In another aspect, a second camera is used (eg, attached to the camera port base). However, VSOM does not require two cameras. In addition to adjusting the xy scanning stage, there is a housing for the stepper motor used for coarse adjustment of the z-axis position of the microscope objective, and a computer-controlled XYZ microcapillary positioner or "robot arm". In many embodiments, the microscope objective is located directly below the microscope "stage plate" (or specimen holder). Various removable stage plates can be inserted into the xy scanning stage depending on the chamber or vessel used in the experiment. Many specimen holders or "stage plates" can be used with the present invention. In some embodiments, non-environmentally controlled specimen containers are used. Chambered coverslips, tissue culture flasks, cell culture dishes, and stage plates for multiwell plates can also be used with the present invention. In some embodiments of the present invention, various sterile specimen containers are simply removed from the cell incubator, mounted on a microscope using a suitable stage plate, and the number of cells in the container is determined by the VSOM system using only transmitted light. Can also be counted. Under this circumstance, cell-by-cell morphological measurements, such as detecting individual cells and splitting them into digital images, are also possible, so that cell morphology, such as shape, area, texture, and nucleus-cytoplasm ratio, is also possible. Features can also be monitored.
[0066]
In some cases, a microenvironmentally controlled cell chamber is physically attached to the stage or connected via a stage plate. In fact, any suitable cell chamber (i.e., a receptacle for containing cells to be analyzed and / or observed). Various types of commercially available environmentally controlled cell chambers are available and are suitable for use with the xy scanning stage. In some embodiments, excess liquid is removed from the specimen container by a peristaltic pump or other pump. In other embodiments, the temperature of the cell chamber (or specimen container) is monitored. In these embodiments, the temperature probe is typically connected to a controller that controls the temperature of the liquid in the chamber or the temperature of the chamber itself. In a preferred embodiment, this non-computer controlled independent temperature control unit is replaced by a computer controllable unit.
[0067]
One additional feature of the cell chamber is the ability to layer a defined gas stream of oxygen components over a specimen container located within the cell chamber. In this manner, a known oxygen tension condition is created in the sample container and the solution, or the CO contained in the gas flowing above the sample container is₂By varying the amount, the pH of the solution containing the bicarbonate buffer can be controlled.
[0068]
In some embodiments, two computer controlled pumps having four independently controllable syringes are positioned adjacent to the optical platform. Each of the two pump units incorporates a daisy-chained electronic controller. In this manner, up to about 99 pump units can be connected, and the two syringes of each pump can be controlled independently.
[0069]
As shown in FIGS. 1 and 2, many of the peripheral devices have a control unit that functions as an interface between the host computer and the robot peripheral device itself. Basic software commands are provided by the manufacturers of these peripherals so that control of the microscope peripherals can be incorporated into custom software running on the host computer. The host computer can be controlled by software running on a remote computer. For example, a coarse Z-axis focus adjustment motor, an xy scan stage, two illumination shutters, a digital camera (and its internal shutter), and an optical filter wheel can all be controlled using this software. In some embodiments, fine control of the focus of the microscope is provided by a piezoelectric positioner screwed into the objective turret of the microscope.
[0070]
1. Microarray format
The present invention provides a continuous assessment of various cell types through a high-throughput screen to more fully assess the physiological responsiveness of cells / individual cells to a particular substance, treatment, or treatment set. Processing provides the ability to rapidly screen large numbers of compounds (eg, drugs or compounds that may have a desired biological activity). In the manipulation and analysis of drug-cell interactions using existing methods, the number of bioassays that can be analyzed in a given amount of time is small, the methods required for compound delivery are complicated, and often large amounts of compounds are tested. Therefore, high-throughput and biologically high-content screening cannot be performed. Thus, a preferred embodiment of the present invention allows for high-throughput screening in a microarray format.
[0071]
In a preferred embodiment of the present invention, optical image signals emitted by individual cells are detected and processed according to the cell type. Thus, placing multiple cell types on a support or plate, identifying the cell type in response to an optical image of the individual cell, and tracking the responsiveness of the individual cell or cell type to an applied environmental stimulus This allows for the creation of non-uniform microarrays. Such cell types may be immobilized at an appropriate low cell density to allow for the detection and analysis of light from a single cell on the aforementioned homogeneous or heterogeneous microarray. In another aspect of the invention, collective light signals from multiple cells that have undergone the same process are detected and measured.
[0072]
In a preferred embodiment, the cells are fixedly attached to a support. Suitable methods for depositing cells in a micropattern array are well known and include, for example, photochemical resist photolithography.
resist-photolithography (Mrksich and Whitesides, Ann. Rev. Biophys. Biomol. Struct. 25: 55-78, 1996). In this photoresist method, a glass plate is uniformly coated with photoresist and a photomask is placed over the photoresist coating to define the desired "array" or pattern. The exposure removes the photoresist in the unmasked areas. The entire photolithographically defined surface is uniformly coated with a hydrophobic material, such as an organosilane, that binds to both the exposed glass areas and the photoresist covered areas. The photoresist is then stripped from the glass surface, exposing an array of exposed glass spots. Next, the glass plate is washed with an organic silane having a hydrophilic or chemically reactive group such as an amino group at the terminal. Hydrophobic organosilanes bind to the exposed glass spots, resulting in a glass plate having an array of hydrophilic or reactive spots (disposed in the area of the original photoresist) on the hydrophobic surface. can get. An array of spots of hydrophilic groups provides a substrate for non-specific and non-covalent binding of specific cells, including those from neurons (Kleinfeld et al., J. Neurosci. 8: 4098-4120; 1988). Similarly, reactive ion etching on silicon wafer surfaces has also been used to obtain surfaces patterned with two different types of textures (Craighead et al., Appl. Phys. Lett. 37: 653, 1980; Craighead et al.). J. Vac. Sci. Technol. 20: 316, 1982; and Suh et al., Proc. SPIE 382: 199, 1983).
[0073]
Another method for producing uniform microarrays of cells is based on specific but non-covalent interactions, and uses a photoresist stamping method to obtain protein-adsorbing alkanethiol-coated gold surfaces. (Singhvi et al., Science)
264: 696-698, 1994). Next, the bare gold surface is coated with an alkanethiol having polyethylene at the end and resistant to protein adsorption. Exposure of the entire surface to laminin, a cell-binding protein found in the extracellular matrix, allows live hepatocytes to adhere uniformly to and proliferate on laminin-coated islets ( Singhvi et al., 1994). Processing that utilizes strong but non-covalent metal chelation has been used to coat gold surfaces with specific protein patterns (Sigal et al., Anal. Chem. 68: 490-497, 1996). In this case, a pattern is formed on the gold surface with alkanethiol having nitriloacetic acid at the end. Bare areas of gold are coated with tri (ethylene glycol) to reduce protein adsorption. Ni²⁺Is observed, the specific adsorption of the five histidine-labeled proteins is kinetically stabilized.
[0074]
More specific uniform cell binding is obtained by chemically cross-linking specific molecules, such as proteins, to reactive sites on patterned substrates (Aplin and Hughes, Analyt. Biochem. 113: 144-148, 1981). In another processing method of substrate patterning, an array of readable spots is created. The glass plate is washed with organosilane, which chemically adsorbs and coats the glass. The organosilane coating is irradiated with deep ultraviolet light through an optical mask that defines the pattern of the array. Irradiation cleaves the Si—C bond to produce reactive Si radicals. Si radicals produce polar silanol groups by reaction with water. As described in U.S. Pat. No. 5,324,591, which is incorporated herein by reference, polar silanol groups constitute an array of spots that can be further modified to spot other reactive molecules. Join. For example, a silane containing a biologically functional group, such as a free amino component, can be reacted with a silanol group. The free amino group can then be used as a covalent attachment site for biological molecules such as proteins, nucleic acids, carbohydrates, and lipids. A method has been described in which a lectin known to interact with the cell surface is covalently attached to a glass substrate via a reactive amino group without patterning (Aplin and Hughes, 1981). ). Optical methods of creating a uniform array of cells on a support require fewer steps and are faster than photoresist methods, but require intense UV light from expensive light sources.
[0075]
In another aspect, the application of the cell array and test stimuli is based on the methods and compositions described in US Pat. No. 6,103,479, which is fully incorporated herein by reference. US Pat. No. 6,103,479 discloses a preferred means of providing a miniaturized cell array capable of delivering a compound of interest (eg, drugs and other biologically active compounds). U.S. Patent No. 6,103,479 discloses means and methods for high-throughput and high-content screening of a cell's physiological response to a biologically active compound. A heterogeneous micropattern array of cells means that each "well" or group of wells on the support has a cell-binding selectivity rather than the cells being distributed in a single, uniform coating on the support surface. Refers to an array of cells on a base distributed in a heterogeneous manner such that they may be unique in plane. Any cell type can be arrayed on a heterogeneous micropattern array of cells, as long as there are molecules in the micropatterned chemical array that can specifically bind to the cell type to be arrayed. Preferred cell types for heterogeneous micropatterned arrays of cells include lymphocytes, cancer cells, neurons, fungi, bacteria, and other prokaryotes and eukaryotes.
[0076]
The method of making a heterogeneous micropattern array of cells of US Pat. No. 6,103,479 comprises preparing a micropattern chemical array, chemically modifying the micropattern chemical array non-uniformly, and Attaching the cells to the heterogeneously modified microchemical pattern array. The base may be glass, plastic, or a silicon wafer, such as a conventional optical microscope cover glass, but may be made of any other material suitable for providing a base. As described herein, the term "well" is used to describe a particular spot on the base and need not be at a particular depth. A method of making such an array is described in U.S. Patent No. 6,103,479. In a preferred embodiment, the modified micropatterned chemical arrays are made in a combinatorial fashion, as described in the patent specification. The resulting wells are heterogeneous (ie, each well or group of wells may be unique in terms of cell binding selectivity). Combinatorial means that the wells are treated variably.
[0077]
2. Microfluidics format
Efficient solution delivery to microarrays of cells attached to solid substrates is facilitated by microfluidic systems. Methods and apparatus for precise handling of small liquid samples are described in Ink Delivery (US Pat. No. 5,233,369; US Pat. No. 5,486,855; US Pat. No. 5,502,467, all of which are hereby incorporated by reference). , Aspiration of biological samples (US Pat. No. 4,982,739, incorporated herein by reference), storage and delivery of reagents (US Pat. No. 5,031,797, And a microelectronic and fluidics compartmentalizer array for clinical diagnostics and chemical synthesis (US Pat. No. 5,585,069, incorporated herein by reference). ) Is disclosed. Further, methods and apparatus for forming microchannels in a solid substrate that can be used to guide small liquid samples along a surface are disclosed (US Pat. No. 5,571,410; US Pat. No. 5,500,071; U.S. Pat. No. 4,344,816, all of which are incorporated herein by reference). A method for delivering a solution to micropatterned living cells as a heterogeneous array on a solid substrate in a closed optical chamber is disclosed in US Pat. No. 6,103,479; see also Patent. As incorporated herein.
[0078]
Preferred embodiments of the heterogeneous micropattern array of cells are as disclosed above. In a preferred embodiment of the liquid delivery system, the chamber is connected to a substrate containing a heterogeneous micropatterned array of cells. The material of the chamber is preferably glass, plastic, or silicon, but can be any other suitable material that can provide a substrate. One embodiment of the chamber has an array of etched areas that match wells in a heterogeneous micropattern array of cells. In addition, microfluidics channels are etched to supply liquid to the etched areas. A series of "waste" channels for removing excess liquid from the etched area may also be connected to the well. The chamber and the micropattern array of cells together constitute a cassette.
[0079]
"Liquid" includes, but is not limited to, a solution of a particular drug, protein, ligand, or other substance that binds to or is taken up by cell surface expressed components. Solutions for reacting with a heterogeneous micropatterned array of cells may also include drug-encapsulated liposomes. In one embodiment, such liposomes are formed from photochemical materials such as photoreactive biosynthetic polymers that release the drug upon exposure (reviewed in Willner and Rubin, Chem. Int. Ed. Engl. 35: 367-385, 1996). ). The drug may be released simultaneously in all channels 14 or may be released sequentially by illuminating individual channels or rows of channels. Controlled release of such agents may be used for kinetic and live cell studies. Control of liquid delivery is possible by a combination of microvalves and micropumps known to those skilled in the art of capillary action (US Pat. No. 5,567,294; US Pat. No. 5,527,673; and US Pat. 5,585,069, all of which are incorporated herein by reference).
[0080]
In a preferred embodiment, the drug or other substance is delivered as follows. Solutions of the substances to be tested for interaction with the cells of the array can be loaded from a 96-well microtiter plate into an array of microcapillary tubes. The array of microcapillary tubes corresponds one-to-one to the microfluidics channels of the chamber, allowing the solution to flow or pump from the microcapillary tubes into the channels. Invert the heterogeneous micropattern array of cells so that the wells are immersed in the liquid-filled etched area. Once the interaction of the liquid with the heterogeneous micropattern array of cells occurs, the light emitted by the heterogeneous micropattern array of cells can be directly measured, and the placement, removal, and processing of the cell array can be automated by robots. It can be converted to optical input for optical guidance.
[0081]
3. Fluorescence format
As described herein, particularly preferred embodiments include fluorescent imaging of cells. Various methods have been developed for imaging fluorescent cells with a microscope and extracting information about the spatial distribution and temporal changes occurring in these cells. Many of these methods and their applications have recently been disclosed (Taylor et al., Am. Scientist 80: 322-335, 1992; see also US Pat. No. 6,103,479). These methods have been designed and optimized to produce a small number of specimens for the purpose of imaging the distribution, quantity, and biochemical environment of fluorescent reporter molecules in cells with high spatial and temporal resolution. It is.
[0082]
Treatment of Cells with Dyes and Fluorescent Reagents and Imaging of Cells (Wang et al., "In Methods in Cell Biology", New
York, Alan R.S. Lis., 29: 1-12, 1989), and genetic manipulation of cells to produce fluorescent proteins such as modified green fluorescent protein (GFP) as reporter molecules are useful detection methods. Aequorea, a kind of jellyfish
Victoria's green fluorescent protein (GFP) has a maximum excitation wavelength at 395 nm, a maximum emission wavelength at 510 nm, and requires no extrinsic factors. Gene expression and protein localization using GFP is described in Chalfy et al., Science.
263: 802-805, 1994. Some properties of wild-type GFP are disclosed by Morise et al. (Biochemistry 13: 2656-2662, 1974) and Ward et al. (Photochem. Photobiol. 31: 611-615, 1980). A paper by Rizzuto et al. (Nature 358: 325-327, 1992) discusses the use of wild-type GFP as a tool for visualizing organelles. Kaeser and Gerdes (FEBS)
Letters 369: 267-271, 1995) reports the visualization of protein trafficking on secretory pathways using wild-type GFP. For the expression of GFP in plant cells, see Hu and Cheng (FEBS).
Letters 369: 331-334, 1995) reported on GFP expression in Drosophila embryos by Davis et al. (Dev. Biology 170: 726-729, 1995). In US Pat. No. 5,491,084, incorporated herein by reference, Aequorea as a reporter molecule fused to another protein of interest.
(Victoria) is disclosed. PCT / DK96 / 00052, incorporated herein by reference, relates to a method for detecting cytologically active substances that affect intracellular processes utilizing a GFP structure having a protein kinase activation site. There is a great deal of literature on GFP proteins in biological systems. For example, PCT / US95 / 10165, incorporated herein by reference, discloses a system that utilizes the expression of a GFP-like protein to separate cells of interest. PCT / GB96 / 00481, which is incorporated herein by reference, discloses the expression of GFP in plants. PCT / US95 / 01425, incorporated herein by reference, discloses the expression of a modified GFP protein in a transformed organism to detect mutagenesis. GFP mutants have been made and used in several biological systems (Hasselhoff et al., Proc. Natl. Acad. Sci. 94: 2122-2127, 1997; Brejc et al., Proc. Natl. Acad Sci. 94: Cheng et al., Nature Biotech. 14: 606-609, 1996; Heim and Tsien, Curr. Biol. 6: 178-192, 1996; and Ehrig et al., FEBS Letters 367: 163-166, 1995. ). U.S. Patent Nos. 5,436,128 and 5,401,629, both of which are incorporated herein by reference, describe receptors comprising transcriptional regulatory elements that are responsive to the activity of cell surface receptors. Disclosed are methods showing assays and configurations for detecting and assessing intracellular conversion of extracellular signals using recombinant cells containing the genetic structure and expressing a cell surface receptor.
[0083]
The ArrayScan ™ system, developed by BioDx, Inc. (U.S. patent application Ser. No. 08 / 810,983), utilizes a luminescently labeled reporter molecule in cells to search for multiple compounds for a particular biological activity. An optical system that determines distribution, environment, or activity. The ArrayScan ™ system includes the steps of providing cells containing luminescent reporter molecules in an array of uniform locations, and scanning a large number of cells at each location with a fluorescent microscope, wherein the scanning involves optical information Is converted to digital data, and the distribution, environment, or activity of the luminescently labeled reporter molecule in the cell is determined using the digital data. Currently used arrays of uniform locations are industry standard 96-well or 384-well microtiter plates. The ArrayScan ™ system includes devices and computerized methods for processing, displaying, and storing data, thereby providing a cell-based high content search in a large-scale microtiter plate format. Increase drug discovery.
[0084]
The foregoing embodiments can be combined to provide methods and apparatus for performing multicolor luminescence reading, microfluidic delivery, and automated environmental control of living cells in a uniform or non-uniform micropattern array.
[0085]
In one aspect, the invention includes a heterogeneous micropatterned array of cells and a method of making the array. The array may include the same cell type that may be treated with another combination of compounds, or may include a combination of cell types that may be treated with one or more compounds. The term "combinatorial" means that a well or group of wells is treated variably. Another aspect of the invention is a heterogeneous micropatterned array of cells, wherein the cells comprise at least one luminescent reporter molecule, and a liquid delivery system for delivering a combination of reagents to the heterogeneous micropatterned array of cells. And a means for detecting, recording, and analyzing a luminescent signal emitted by a luminescent reporter molecule. In another aspect of the invention, a luminescent reader for detecting luminescent signals emitted by luminescent reporter molecules in a heterogeneous micropatterned array of cells, a digital detector for receiving data from the luminescent reader, and a light detector And a computer means for receiving and processing the digital data sent from the vessel.
[0086]
In another embodiment, the cells are modified with a fluorescent indicator that represents the chemical or molecular properties of the cells, seeded on a heterogeneous micropatterned chemical array, and analyzed live. Examples of such indicators are described in Giuilano et al., Ann. Rev .. Biophys. Biomol. Struct. 24: 405-434, 1995; Harootunian et al., Mol. Biol. Cell 4: 993-1002, 1993; Post et al., Mol. Biol. Cell 6: 1755-1768, 1995; Gonzalez and Tsien, Biophys. J. 69: 1272-1280, 1995; Swamithanan et al., Biophys. J. 72: 1900-1907, 1997; and Chalfy et al., Science 263: 802-805, 1994. The indicator may be introduced into the cells by one or a combination of various physical methods before or after seeding the cells on the array, such as diffusion through cell membranes ( Haugland, Handbook of Fluorescent Probes and Research Chemicals (Handbook
of Fluorescent probes and research chemicals), separate volume 6, edited by Molecular Probes, Eugene, 1996), mechanical perturbation of cell membranes (McNeil et al., J. Cell Biology 98: 15615-64; Cell Science 102: 533-541, 1992; Clarke et al., BioTechniques 17: 1118-1125, 1994), or genetic manipulations such that the indicator is expressed in cells under defined conditions (Chalfy et al., 1994). However, the present invention is not limited to these. In a preferred embodiment, the cells contain a luminescent reporter gene, although other types of reporter genes are suitable, including genes encoding chemiluminescent proteins. Living cell studies allow one to analyze the physiological state of cells based on luminescence during the life cycle or upon contact with drugs or other reactive substances.
[0087]
In another aspect of the invention, a heterogeneous micropatterned array of cells can be created wherein the cells comprise at least one luminescent reporter molecule, and the reagent can be delivered to the heterogeneous micropatterned array of cells. Performing high-throughput searches by contacting a heterogeneous micropattern array of cells with a liquid delivery system to obtain a luminescent image of the individual cells at the appropriate magnification to obtain individual images of the cells And analyzing the individual cells. Next, using a set of luminescent reagents having different physiological and spectral properties, scanning an array of individual cells to obtain luminescent signals emitted by luminescent reporter molecules in the cells, The step of converting to digital data and the step of utilizing the digital data to determine the distribution, environment, or activity of the luminescent reporter molecule in individual cells provides for high content detection in the reacted wells.
[0088]
A cassette containing a heterogeneous micropattern array of individual cells and a chamber is inserted into a VSOM optical reader. Optical readers control cassette handling, the environment (eg, the temperature that is important for living cells), control of solution delivery to wells, and from arrays of cells for one or more cells at a time. Opto-mechanical device for analyzing emitted luminescence. The instrument automatically controls the addition of the solution based on the light signals and algorithms emitted by the individual cells. In a preferred embodiment, the optical reader includes an integrated circuit inspection station using a fluorescence microscope as a reader and a micro robot for scanning the cassette. The cassette is held in a storage compartment and is retrieved from the storage compartment by a computer controlled robotic arm. The robot arm inserts the cassette into the luminescence reading device. Another robotic arm removes the cassette from the light emitting reader and places it in the second storage compartment.
[0089]
Systems integrating environmental controls, microrobots and optical readers are manufactured by companies such as Carl Zeiss [Jena, GmbH]. In addition to facilitating robotic handling, liquid delivery, and fast and precise scanning, two reading modes of high content and high throughput are provided. High content readings are essentially the same as readings with the ArrayScan reader (US patent application Ser. No. 08 / 810,983). In the high content mode, each location of the heterogeneous micropattern array of cells is imaged at a magnification of 5x to 40x or more, and a sufficient number of fields are recorded to make the desired statistical resolution measurements.
[0090]
In one embodiment, the optical reader is an opto-mechanical design that is an upright or inverted fluorescent microscope with a computer controlled x, y, z stage, a low magnification (eg, 0.5 ×) objective, and one Or includes a computer-controlled rotating nosepiece holding a plurality of higher magnification objectives, a white light source having an excitation filter wheel, a dichroic filter system having an emission filter, and a detector (eg, a cooled charge coupled device). . In another aspect, the optical reader may utilize a scanning laser beam in a confocal or standard illumination mode. The spectrum was measured by Carl Zeiss (Carl
Zeiss) (Jena, GmbH, Germany) or as reported by Denk et al. (Science 248: 73, 1990) based on multiple laser lines or groups of individual laser diodes. Selected. Another aspect of the high-throughput search mode is a low-resolution system having an array of luminescence exciters (1 × 8, 1 × 12, etc.) and a luminescence emission detector, wherein the wells of a heterogeneous micropattern array of cells are provided. A system that scans a subset of is used. The system has a fiber optic bundle in a preferred embodiment, but may be any system that induces luminescence excitation light and collects luminescence radiation emitted by the same well.
[0091]
B. Local host computer (or client)
The computer controlled peripherals may be independently controlled by the local computer, or the local computer may function as a conduit for instructions issued by a remote computer. A typical "instrument" or remote control of a microscope, or a VSOM system on the Internet, has several advantages, as described below. For example, a central computer may have higher processing and archiving capabilities, as well as more advanced databases and other software packages, thereby bringing these advantages closer to individual optical platforms. This eliminates the need for costly duplication on multiple local computers.
[0092]
C. Remote computer, network software, database, informatics, and analysis algorithm for VSOM control
From a computer science perspective, the present invention required various components. Particularly preferred embodiments include the following components: (1) computer control of the microscope peripherals, (2) segmentation of cells imaged on single or multiple channels, (3) used in certain experiments. Commentary on recipes, (4) automated analysis of cellular responses as a function of applied stimulus, (5) off-line correlation, clustering, and learning techniques that compare differences between two cell types as a function of assay And (6) online learning techniques that dynamically control and infer policies (ie, protocols or recipes) to derive a direct response from two or more different cell types. The interaction between the various system components is shown in FIGS. In a particularly preferred embodiment, the complete integration of all system components includes a specially designed graphical user interface (GUI), online image analysis software, database integration for offline analysis, and efficient cell response within the model system. Visualization, as well as offline and online learning techniques (eg, algorithms).
[0093]
In yet another aspect, the invention provides an image bioinformatics system (ie, "BioSig") for visual servo controlled light microscopy. These aspects of the invention provide the basis for cataloging cellular responses as a function of specific conditions and stimuli for VSOM experiments. The present invention provides a system architecture, a functional data model for representing experimental protocols, novel algorithms for image analysis, and statistical analysis. The architecture provides for remote sharing operation of at least one VSOM device, and couples device operation with image acquisition and captioning. In a preferred embodiment, the information is stored in an object-oriented database that can permanently store objects and their associations. In a particularly preferred embodiment, the algorithm extracts physiological information from individual cells at the subcellular level and maps it to cellular responses and ultimate biological consequences.
[0094]
The VSOM system, in some preferred embodiments, periodically acquires images, segments those images, measures cell responses on a field of view containing hundreds of cells, and records those responses in an archive system. Use a workflow model. These reactions allow for the directed measurement and tracking of individual cell responses. You can then browse the archive system and query the archive via a web-based interface. Example of browsing raw data, segmentation results, and corresponding metadata (eg, response calculations over 2.5 hours).
[0095]
One of the significant issues overcome by the present invention is related to the fact that biological experiments require testing with large populations and correlate distant features with annotation data. . The present invention relates to an image bioinformatics system for integrated image acquisition, captioning, and hierarchical image extraction to create a database that records location and expression information for multiple targets along with location references and morphological features I will provide a. Statistical and visualization tools are integrated to enable hypothesis testing and data mining. This is achieved by augmenting and extending the infrastructure for telemicroscopy development (see, eg, Parvin et al., “Discussion: Channels for Distributed Microscopy and Informatics (Deep).
View: A Channel for Distributed Microscopy and Information, "IEEE High Performance Computing and Networking Conference (IEEE).
Conf. on High Performance Computing and Networking [1999], and Parvin et al., "Declarative Flow Control for Distributed Devices.
Flow Control for Distributed Instrumentation ”, IEEE Symposium on Grid Cluster Computing (IEEE)
Int. Symposium on the Grid and Cluster Computing), May 2001). The present invention provides a variety of novel components, including but not limited to the following: distributed architecture for image acquisition, structure and function at the subcellular level. Novel algorithms for characterization, archiving of each cell's response to specific stimuli, and use of response calculations to drive the instrument to specific states.
[0096]
D. Informatics
In a particularly preferred embodiment, the informatics system includes a data model, a presentation manager, and a query manager. These subsystems are decoupled to facilitate development, testing, and maintenance. The purpose of the data model component is to obtain the required annotation data and to link them with the computed image representations that represent the hypothetical signals and response pathways. The model is object-oriented and allows two-way tracking of the annotation and the measured feature data. The presentation manager includes at least two mappings, including mappings between the data model and the user interface, so that user interface wiring is avoided despite changes in the underlying data model, and the interface configured at runtime is more flexible. Provides two distinct features. In addition, the ability to display specific queries in text or graphics is provided. The query manager maps high-level user queries to Java objects that execute the data model. Thus, in a preferred embodiment, the present invention simplifies and hides the detailed manipulation of the database from the end user.
[0097]
1. Data model
The data model of the embodiment shown in FIG. 6 panel A is object-oriented, linking a particular project to features calculated from a collection of images. Links are bi-directional, allowing information to be tracked from any endpoint. Each project has its own database linked to the test.
[0098]
2. Presentation Manager
The presentation manager has two main features: browsing the database and visualizing the results of the query function. Browsing the database is performed against a predefined schema that obtains the annotation data, images, and corresponding features. The data model shown in FIG. 6 panel A is represented in XSD (XML scheme), and the presentation manager uses this representation and the corresponding stylesheet (XSL) to construct views in the database for viewing and updating. In this manner, the GUI wiring is bypassed to favor a more flexible and dynamically generated user interface. In general, such mappings can create complex implementation problems. However, the present invention simplifies the presentation system and allows one layer to be viewed and updated simultaneously. “Layer” refers to the navigation between objects linked by association, aggregation, and inheritance from other objects.
[0099]
In one embodiment, each scientific experiment may include up to a number of images corresponding to a population of cells at a particular point in time. In many cases, at each point in time, it is necessary to visualize a collection of images with several (2-3) images characterizing the average behavior of the set. Average behavior is often described in terms of particular characteristics. According to the present invention, it is possible to construct a series of images in which each image corresponds to a representative of a set of images at a certain time. In addition, the presentation manager can display the results of the query function in text or graphics. Graphics include dose-response plots and scatter plots that display compound characteristics as a function of the independent variables.
[0100]
3. Query Manager
The query manager provides a predefined set of operators to assist with visualization and hypothesis testing. These operators assist in delineating the contrast between the calculated features and the corresponding annotation data, and perform various statistical methods such as analysis of variance and principal components analysis (PCA). In some embodiments, these operators allow the cellular response to be interpreted for final model reconstruction. In an object-oriented database, measurements, such as analysis of variance, are simplified because each computed feature must be mapped to its source (eg, cell culture or animal). One example of such a high level operator is the correlation of certain calculated features to independent variables. The present invention provides a visual query interface for mapping user queries to database operations, calculating results, and transferring the results to a display manager. In some aspects, the actual calculation may include analysis of variance for display purposes (eg, to associate a particular measurement with multiple independent samples) or PCA (eg, reduction of the calculated feature vectors). ) Is included.
[0101]
4. Nuclear extraction
Automated delineation of subcompartments of cells is an important step in monitoring the response of cells to applied stimuli. In some cases, the cells of interest may touch or overlap each other and thus be difficult to segment. Conventional approaches utilized during the development of the present invention (eg, Cong and Parvin, Patt. Recog., 33: 1383-1393 [2000]; and Parvin et al., "Biosig: for the study of the mechanism of intracellular signaling. Bioinformatics system (Biosig: A Bioinformatic System for Studying the Mechanism of Inter-Cell Signaling), IEEE
Inter. Sympos. Bio-Informatics Biomed. Engineer. , Pages 281-288 [2000]), both step edges and roof edges were used to partition nuclear agglomerates in a globally consistent manner. The step edge corresponds to the boundary between the nucleus and the background, and the roof edge corresponds to the boundary between adjacent nuclei. One of the unique features of this system was the hyperquadratic representation of each hypothesis and the use of this representation for global consistency. Global consistency was obtained by a cost function minimized by dynamic programming. The present invention provides a simpler and more powerful approach. The main difficulty with previous approaches was that the detection and grouping of wrinkle boundaries increased the complexity of the system and, consequently, reduced its reliability. The invention is also model-based in that it assumes that the projection of each kernel in image space is quadratic. However, in some preferred aspects of the invention, instead of grouping step edges and roof edges, one starts with a representation corresponding to the zero crossings of the image. Next, the zero-crossing image is filtered with geometric and illumination constraints to generate a binarized agglomerate of nuclei. Next, the nuclear agglomerate is partitioned into a plurality of nuclei by a process called “center of gravity conversion”. Each stage of the calculation protocol is shown in FIG. The centroid transformation essentially projects each point on the contour to a localized center of mass. As shown in FIG. 7 panel B, solution normalization is performed to remove noise and other artifacts on the contour.
[0102]
a. Edge detection
Edge localization is based on the zero crossings of the local coordinate system. As explained in detail below, these edges correspond to minimizing the upper bound.
[0103]
The majority of normalization techniques for smoothing are:_R| ∇f |²It is based on minimization of integrals. However, this equation has no control over the local properties of f. In other words, even when the global average of | ∇f | is small, f may locally change suddenly. One way to overcome this problem is to minimize | ∇f | at all points, thereby minimizing the upper limit of | ∇f |. Therefore, the functional H (f) = sup_R | ∇f | can be considered to be the “limit” of a series of functionals (Equation 1) of the following equation:
(Equation 1)

Functional H_NEuler's equation that minimizes (f) is expressed by the following equation (Equation 2).
(Equation 2)

In the formulas, subscripts represent derivatives, for example,
(Equation 3)

Equation 3 is obtained by eliminating the first coefficient and making N → ∞.
(Equation 4)

Equation (3) is called the Infinite Laplace Equation (ILE) and has been studied extensively (eg, Aronsson, Arikv. Materialmatik, 6: 551-561 [1966]; Aronsson, Arikv. Materialmatik, 7: 133-151). Aronsson, Manusscripta Math., 41: 133-151 [1981]; and Evans, Electron. J. Different. Equat., [Http://ejde.math.swt.edu/Volumes/1993/. -Evans / abstr.html] 3: 1-9 [1993]). Some important properties of equation (2) are that at least one solution exists, and C if the solution is redefined in a reasonably weak sense¹That there is a continuous solution of. Further, the locus of the gradient of f is a convex curve or a straight line, and there is no stop point | 停 f | = 0 in R. In a local coordinate system in which the local coordinate system is defined by ∇f, equation (3) has a zero crossing (f_xx+ F_yy= 0) is interesting.
[0104]
Applying equation (3) to the raw data results in a more stable (ie, no leakage between foreground and background) zero-crossing map than Laplace form. This zero-crossing map has a closed contour with erroneous regions that can be easily filtered under intensity and size constraints. The filtered segment image is then used for centroid transformation.
[0105]
b. Tensor-based feature analysis
Vector field analysis is a well studied technique in computer vision and pattern recognition as well as classical fields such as mathematics and physics. Numerous concepts and tools have been developed for exercise-based applications. The present invention extends the current state of the art to the segmentation and interpretation of scientific images. Intensity image f₀Given (x, y), there is a natural vector field derived from the intensity image, which corresponds to a tilt field with a rotation of π / 2:
(Equation 5)

The problem here is that the vector field "v" is noisy and requires some kind of normalization to be introduced. This is expressed as:
(Equation 6)

Where u is the normalized vector field. Next, an elliptic PDE solution is obtained by an iterative method. A first-order approximation can be determined by a simple linear (Gaussian) scale space model of the following equation (Lindberg, J. Appl. Stat, pages 225-270 [1994]):
(Equation 7)

Where:
(Equation 8)

Is the standard deviation
(Equation 9)

Which is the Gauss kernel.
[0106]
Second, the singularities of the vector field give a compact abstraction of the dense vector representation. These singularities can be characterized by point or linear features. Point features include velocity and saddle point. After these patterns are detected, the corresponding objects can be extracted. Rao and Jain (IEEE
Trans. Patt. Anal. Mach. Intel. , Pages 225-270 [1994]) detected singularities by using local Jacobian for feature extraction from the underlying vector field. However, the approach of the present invention is more powerful and allows vortex size measurements. The vortex size complements the localization of the convex blobs, even when the convex blobs are connected or in contact with each other. It is based on the Jordan Index, which identifies the location of the singular point in the underlying vector field. Let F = (u, v) be a vector field, and let J be a Jordan curve without a critical point. The index of J is defined by:
(Equation 10)

At each point P, a small circle (J^R _P) Is selected, and Index (J^R _P) Is calculated. Next, the flow fields (u, v) can be classified as follows:
1. If the vortex index is equal to +1 (ie, the singularities in the vector field are given by Rao and Jain above), and
2. The saddle point index is equal to -1.
There are no nodes in the vector field because the divergence of the vector field is zero at all points as follows:
(Equation 11)

Next, the size of the vortex is estimated by a simple search technique. If a point a (x, y) is a vortex, its size R * (x, y) is defined by:
(Equation 12)

In other words, R * (x, y) is J^R _{(X, y)} Is the largest R such that the index of R remains at 1. The Jordan index, which is the integral of the first partial derivative, is in some sense a function of the intensity image. In contrast, other techniques are based on higher-order derivatives that are more prone to noise. Various figures show the results of vortex and region segmentation in the synthesized images, nuclei labeled with fluorescent dyes, and cells observed with transmitted light.
[0107]
FIG. 41 panels C and D show the nuclei results of live and lysed cells labeled with the fluorescent dye Hoechst 33342. FIG. 41 Panels E and F show the results with live cells imaged using a 20 × phase contrast objective and transmitted light. The latter technique is important because conventional techniques for locating individual cells are limited to fluorography, which increases the complexity of designing dynamic experiments. Of the digital images shown in this specification, only panels E and F in FIG. 41 were taken without using a 10 × NA0.5 Zeiss Fluar objective. All images were 1024 × 1024 and the pixels of the CCD chip in the camera were 6.8 × 6.8μ. Therefore, in all images taken at 10 ×, the field of view is approximately 700 μ in height and 700 μ in width.
[0108]
c. Normalized centroid transformation
Let I (x, y) be the original image. Each point (x₀, Y₀)), The contour is defined by the following equation (Equation 4):
(Equation 13)

Extending and truncating the above equation using the Taylor series gives the following equation (Equation 5):
[Equation 14]

In the above equation, u = xx₀And v = yy₀Or
(Equation 15)

Is the Hessian matrix in the standard form (Equation 6)
(Equation 16)

At
[Equation 17]

Is the gradient of the intensity and w = (u, v)^TIs the center of gravity in the local coordinate system. It is well known that the center of gravity of the quadratic curve defined by equation (6) satisfies the linear constraint of equation (7):
(Equation 18)

If A is not a singular point, the center of gravity can be determined directly (ie, the following equation (Equation 8)):
[Equation 19]

However, this is not always true, and in general the zero set defined by the following equation (Equation 9)
(Equation 20)

Is non-trivial. It can be further divided into the following two categories:
1. Uniform areas corresponding to areas where the intensity gradient is zero. For binary images, the information is only on the contour; and
2. Non-uniform areas corresponding to elliptical shapes in gray level images.
[0109]
The centroid of these points is not well defined. Further, from the viewpoint of calculation stability, these close points cannot be calculated with high reliability because of singularities. To address these difficulties, we need to normalize the problem. Let the center of gravity of (x, y) be (u (x, y),
v (x, y))^T, The normalized model can be represented by the following Equation 10 or Equation 11:
(Equation 21)

Or
(Equation 22)

In the equation, the first and second terms are estimation errors, the third term is the smoothness of the limiting condition, and α (> 0) is the weight. The solution to this problem is called "normalized centroid transformation" (RCT). The Euler-Lagrange equation of Equation 11, which is a variational problem, is:
(Equation 23)

[0110]
When the difference approximation of the partial derivative is substituted into the above difference equation, the following equation (Equation 13) is obtained.
[Equation 24]

The above equation can be rewritten as the following equation (Equation 14).
(Equation 25)

Where (Equation 15)
(Equation 26)

It is.
These coefficients are all functions of partial derivatives and are near (u (x, y), v (x, y)). Determinant (Equation 16):
[Equation 27]

Is always positive, and the solution of equation (14) becomes the following equation (equation 17).
[Equation 28]

Therefore, the estimated partial derivative and the previous estimate (uⁿ, Vⁿ), A new set of estimates (u^{n + 1}, V^{n + 1}) Becomes the following equation (Equation 18).
(Equation 29)

[0111]
e. Expression and classification
Live cell imaging, and generally fluorescence microscopy imaging, is often multispectral to separate structural and functional information. In some embodiments, the sample is labeled with a fluorescent dye and nuclear information (eg, shape and composition) is obtained by imaging at 360 nm. The reaction is imaged at other excitation wavelengths (eg, 490 nm and 570 nm). In some embodiments of the invention, each nucleus is represented by an ellipse and a hyperquadratic function, and the response is read directly in other channels.
[0112]
The elliptical fitting is 4ac-b²Polynomial F (a, x) = ax subject to the constraint condition = 1²+ Bxy + cy²+ Dx + ey + f (Fitzgibbon et al., Proc. Intl. Conf. Patt. Recogn., 253-257 [1996]).
[0113]
Superquadratic functions of 2D (Hanson, Comp. Vision Graph. Image Proc., 44: 191-210 [1988]; and Kumar et al., IEEE).
Trans. Patt. Anal. Mach. Intel. , 17: 1079-1083 [1995]) is a closed curve defined by the following equation (Equation 19).
[Equation 30]

γ_i > 0 (Roskelley et al., Curr. Opin. Cell Biol., 7: 736-747 [1995]), yielding the following equation (Equation 20).
[Equation 31]

The above equation corresponds to each i parallel line segment pair. These line segments define (for large γ) a convex polytope in which the superquadric is restricted. This representation is valid for a wide range of shapes that need not be symmetric. Parameter A_iAnd B_iDetermines the slope of the bond line and the distance γ between these two_iAnd C_iDetermines the "squareness" of the shape.
[0114]
The fitting problem is as follows. n segments
(Equation 32)

M data points P_j= (X_j, Y_j), J = 1, 2,. . . , M is given. The cost function is defined by the following equation (Equation 21):
[Equation 33]

Where:
[Equation 34]

∇ is a gradient operator, λ is a normalization parameter, Q_iIs a constraint term (Kumar et al., IEEE Trans. Patt. Anal. Mach. Intell., 17: 1079-1083 [1995]). Parameter A_i, B_i, C_i, And γ_iAre based on the Levenberg-Marquarr nonlinear optimization method (Press et al., Numerical Recipe in C).
Recipes in C) ", Cambridge University Press [1992]), by minimizing ε from appropriate initial guesses (Kumar et al., IEEE).
Trans. Patt. Anal. Mach. Intel. 17: 1079-1083 [1995]). Each nucleus in the image is further classified by its location in the lumen. FIG. 8 shows an example of ellipse fitting and classification of nuclei in an image.
[0115]
f. Analysis of normalized centroid transformation
In one aspect of the invention, the extraction of nuclei in the screen is different from known morphological operators. One specific technique is a watershed transform that views the image as a three-dimensional structure (eg, Najman and Schmidt, IEEE).
Trans. Patt. Anal. Mach. Intel. , 18: 1163-1173 [1996]). By flooding the image from its local minima and suppressing the merging of regions from different local minima, the image is captured and basin-bounded, as well as watersheds.
line). In another aspect, such partitioning is initiated from an edge by finding a downstream path from each edge point to a local minimum. Watershed transformations lead to over-segmentation, which requires complex morphological operations to merge adjacent regions. In contrast, the preferred embodiment of the present invention starts with the binarized image and subsequently collapses the boundary points belonging to the nucleus into a single point. However, from the viewpoint of moving boundary points to local basins to clarify the natural division, the watershed transformation and the centroid transformation have the same idea. Using this method, the energy landscape corresponding to the distance transformation shows a number of local minima. In contrast, the normalized centroid transform has a smooth energy landscape with a single local minimum.
[0116]
The centroid transformation essentially divides the binary blob into separate convex regions, with each convex region corresponding to a nucleus or other subsection. Partitioning is evolutionary and occurs along natural boundaries. In contrast, in earlier studies performed during the development process of the present invention (see, eg, Cong and Parvin, Patt. Recog., 33: 1383-1393 [2000]), such partitioning is not necessarily the subject's natural Did not occur along the border.
[0117]
Thus, the present invention provides a more detailed picture of the signals generated between cells as a result of various stimuli (eg, exogenously applied stimuli) that elicit a biological response (eg, a biological function). Methods for bioinformatics approaches to microscopy and image analysis useful for: These methods are described in more detail below.
[0118]
Importantly, the VSOM of the present invention is suitable for interfacing with DeepView, a channel for distributed microscopy and informatics that has been reported in more detail in our previous work. . Deep View is a scalable interface for adding any device to its framework. Deep
View also provides a means for transferring and viewing data over wide area networks. The system has a generic GUI that interacts with the targeted device via its characteristic advertisement. For VSOM, the relevant properties are computer controlled syringe, shutter, and XY stage, Z-axis focus motor, filter wheel, and camera control parameters. This concept of device advertising performance (functionality) provides a more uniform interface and reduces maintenance costs by allowing software to be reusable.
[0119]
E. FIG. Segmentation
As noted above, automated depiction of nuclei and cytoplasm is an important step in mapping cellular responses to specific cellular compartments. The computing component interacts with the VSOM system to obtain image data from one or more channels (eg, by rotating a filter wheel or opening and closing a shutter). The segmentation of an image obtained from a single channel is as described above. The outline of multi-channel segmentation will be described below.
[0120]
In the analysis of multi-channel images, another type of data collection occurs, which takes a long time. In this scenario, the cells are imaged using different fluorescent probes and using different excitation filters to highlight different components of the cells. In this case, the system must be capable of simultaneous analysis of multiple channels of data, selection of specific intracellular elements, and repeated measurements. An efficient approach for fusion based on the Bayesian framework has been developed. As shown in studies performed during the development of the present invention (Parvin et al., AAAI
Symp. Appln. Comp. Vis. Med. Imaging [1995]), this approach can be extended to 3D (using confocal microscopy). The purpose of data fusion is to complement the different modalities for segmentation and labeling. From this point of view, the segmentation procedure requires that each pixel in the data area be labeled accordingly. However, there are many ambiguities that can complicate the labeling process. These ambiguities can arise from local processing only and due to the absence of high level feedback.
[0121]
Sources of ambiguity include data corruption due to noise, limitations in the performance of algorithms that extract local features, and the presence of unnecessary features that hinder the tagging task. In one aspect of region segmentation, an estimation of the average intensity of each region is achieved, which is achieved by analyzing the histogram. In some aspects, the initial location of the peak in the histogram is approximated and then refined by a least squares approximation. The next step in the calculation process is to use these peaks as cluster centers to increase local consistency in image space. At this time, a Bayesian framework is used to label images based on multi-channel information. Specifically, the present invention uses a Bayesian hierarchical model with three levels. The first level is a basic Class Z model. Under ideal conditions, viewing Z in M different modalities (eg, three fluorescent images, each corresponding to blue, green, and red fluorescence, respectively) will result in corruption due to various noise in the imaging system. Ideal expression X of the data_iIs specified. The third level of the hierarchy is data Y_iCorrespond to the actual observations. The first level uses a Markov random field (MRF) with a prior probability density function in the form of an Ising model with positive parameters. This parameter facilitates cooperation between neighboring pixels. In the Bayesian framework, an assumption may be made that there is prior ignorance of the content of the screen. The second level models X as a conditional dependent variable given Z (ie, P (X₁,. . . , X_MZ) = P (X₁Z). . . P (X_MZ). This is reasonable if the ideal response of the green channel (ie, cytoplasm) and the blue channel (ie, nucleus) do not affect each other if one classification is labeled as cytoplasm. The third level of the hierarchy, Y_iIs modeled as a Gaussian distribution. As a result, Y_iIs X_iIs a locally blurred expression. The complete model is expressed as:
(Equation 35)

Obtain Xi and Z using maximum posterior probability (MAP) estimation. However, since the MAP calculation is inherently infeasible, iterative condition mode (iterative
Find local optimal conditions using numerical approximations based on conditioning modes (ICM) (see Besag, J. Roy. Stat. Soc. B, 48: 259-302 [1986]).
[0122]
F. software
As described in detail herein, the VSOM functional architecture used in the experiments described herein consists of an image acquisition module, an image analysis module, a servo control, and an archive. The image acquisition module uses six excitation filters to provide a means of time-lapse high resolution video microscopy. This module has a recipe manager that allows manual or pre-programmed image acquisition of various time and excitation frequencies. Recipes are expressed in XML notation that provides semantic interoperability. The analysis module integrates a unique segmentation algorithm that provides a feature-based summary of the image based on size, location, response, and boundary contour. Thus, the analysis module has a pool of unique segmentation algorithms that provide a feature-based summary of the image based on the attributed graph model. As a result, each node in the graph corresponds to a homogeneous region (eg, nucleus, cytoplasm, etc.) in the image.
[0123]
The attributes of each node include the boundary contour, the parametric representation of this contour using a hyperquadratic function, the derived features, and the response of the subcellular compartment under observation. The links in the graph code the adjacencies between the various nodes. The attributed graph model fully represents the structural definition of each cell and its neighbors. The servo control module controls the concentration of various compounds in the tissue culture vessel by controlling the flow of any of the four syringes located near the microscope. Servo control provides three modes of operation, each mode being registered sequentially in the recipe manager. These modes include: (1) a static recipe, in which the flow rate of each compound, its start time, and duration are specified; (2) the flow rate as a function of the specific response of the cells under observation; A dynamic recipe mode, varying duration; and (3) a modulated recipe, controlled by a program that turns the flow on and off at a specified frequency. The archiving system stores the images, their calculated graph models, and the caption data in a flat file environment. One of the key features of the present invention is that it can be remotely operated from multiple locations. This unique feature allows researchers to collaborate on a given experiment, even when physically located at geographically separated locations.
[0124]
The present invention maximizes the flexibility available to the user. For example, the user enters a recipe for a particular experiment that allows control of the perfusion rate into the cell chamber, setting of the camera exposure time, selection of the position of the optical filter wheel, and setting of the sampling rate for data collection. be able to. Herein, one preferred embodiment of the operable VSOM software package for image acquisition and control is referred to as "VX5" and another preferred embodiment is referred to as "ADR". VX5 is designed for static recipe control of the system, and ADR is designed for performing visual servo control experiments of dynamic recipes locally or over a network. FIG.
Schematic view of View's functional architecture (Panel A), sample XML recipe file for VX5 execution (Panel C in FIG. 14), example data in pump log (Panel B in FIG. 14), and example ICS digital image header (FIG. 14 panel D) is shown.
[0125]
G. FIG. Saving feature-based images
Currently, archival models are typically built on flat files, which are difficult to retrieve and maintain. The present invention provides a means to utilize an active bioinformatics program to build the schema required for storage handling. FIG. 15 shows the design of one schema. This design is a feature-based spatio-temporal database that is useful for adapting to VSOM and building models of how specific subcellular compartments have “responded” to applied stimuli and to what extent. It has been extended to create The multidimensional database shown in FIG. 16 provides a representation of the cell response, including the response curves of each subcellular compartment, and the transfer of the fluorescent probe from one compartment to the next.
[0126]
During use of the preferred embodiment, each experiment includes a target frame (ie, an image) corresponding to a particular physical location in a specimen container (eg, a Petri dish). The intracellular structure is segmented and stored as attribute graphs along with the annotation data. The target frame is then observed as a series of digital images collected over time while applying stimuli to the living cells under computer control. The response of each compartment is recorded, and the movement of the fluorescent probe moving into or between cells is tracked.
[0127]
The functional view of stored data is hierarchical. At the lowest level, the raw image and the corresponding physical annotation of the image are stored. The next level stores the attributed graph model (ie, structure), the corresponding response (ie, function), and the dynamic remarks (eg, stimulus type, flow rate, concentration, etc.) of environmental conditions. The next level summarizes the temporal expression of each compartment within the cell as a response curve, along with the trajectory of the compound as it redistributes among the various intracellular compartments. These responses need not be similar for all cells under observation. However, these reactions are expected to generate clusters such that each cluster corresponds to a particular type of cell population or a population of cells in a similar physiological state. These clusters must be identified for subsequent correlation studies. Each cluster corresponds to a time series event, which corresponds to a particular subcellular compartment of the cell. The average behavior of these reactions in each cluster is represented by a single value decomposition. Formally, the observed response is a multi-valued function defined by the following equation (Equation 28):
[Equation 36]

Where f is the average fluorescence intensity observed in a particular section at time t, [Dye] is the concentration of the dye, Channel is the position of the excitation filter, T is the temperature, and Flow is the change in the flow of the injected compound. Corresponding to In order to store observation results in a database for subsequent knowledge discovery, a schema for acquiring this time-varying information is developed. Each experimental cycle generates a fingerprint for each cell observed. From these individual responses, a compact representation of the intracellular activity in the image space is obtained, which can be visualized on a cell-by-cell, cell-population, cell-type, etc.
[0128]
H. Learning from cellular reactions
The learning techniques of the present invention are used to optimize and discover cell-specific differences between cell types. In this context, the calculated response and its tissue structure and trajectory assist in the development of a model to facilitate subsequent step-by-step searching for an optimized cell type identification. The basis of any learning approach is policy, reward, value, and a model of the environment. One of the goals of model reconstruction is to evaluate various policies and reward functions. The model is expected to be different for one cell type and another. These variations are significant for their own merits and can be used as valuable validations if quantified. The database content provides a basis for evaluation, refinement, and localization of similarity measures. Similarity is measured by analyzing clusters based on state and behavior. These clusters assist in calculating co-occurrence statistics for various reactions to establish equivalence classes. Alternatively, an expression in a certain sequence is generated by a cell reaction and its corresponding attribute (dynamic commentary). This representation is then used to build co-occurrence statistics in the sequence along with transition probabilities. Co-occurrence statistics assist in setting hypotheses for binary tree classification of representative feature sets. This method is useful, in part, for the detection and classification of correlated sequences. Further, once the model is established, means for finding outliers can be obtained by co-occurrence statistics. The transition probabilities provide knowledge of the environment, which can then be used for reinforcement learning. Further, a means is provided for modeling cellular responses as a linear system over time. Linear systems are a simple and powerful way to characterize the behavior of the system. In this context, a state-space representation of the system with variable memory (delay) is constructed, and the parameters of the associated matrix are estimated and verified.
[0129]
I. Reinforcement learning
In this phase, known knowledge is used as a starting point to learn how to vary the uptake of the fluorescent probe and optimize the uptake in cells in a systematic way. This is commonly referred to as "reinforcement learning" and is, in some aspects, modeled as a Markov decision process (MDP). Certain MDPs alter their state (ie, the observed response) and alter the intracellular concentration of the fluorescent probe (eg, the concentration of physiologically important ions in the medium that inject biological inhibitors). , Etc.). The reward in such a system can be measured by a change in the calculated response or a change in its episodic slope. Further, in some embodiments, a transition probability (of MDP) is then calculated based on observations obtained from the same assay in the database. These transition probabilities provide a technique that allows for an optimal policy (ie, schedule). Such policies can generally be expressed as dynamic programming, Monte Carlo, and time difference learning (TD learning). Dynamic programming requires a model of the cellular response. Such a model is hypothesized from a database. In addition, model accessibility increases as the database grows. The Monte Carlo method does not require a model, but is not suitable for step-by-step incremental calculations. On the other hand, time difference learning does not require a model and is completely incremental. However, all three methods are applicable to aspects of the present invention. For example, if there is a sample transition probability (not perfect), Monte Carlo can be used in one incremental step (ie, one episode). Also, the time difference method is a combination of Monte Carlo and dynamic programming (ie, learning directly from raw experience without a model, and then updating the estimate based in part on the learned estimate).
[0130]
J. A method for rapidly discovering physiological features that distinguish malignant and non-malignant breast epithelial cells
Recent advances in in vitro growth of primary breast tumor cells have made it possible to passage and grow a relatively small number of tumor cells harvested by fine needle aspiration (FNA) in culture (Li et al., Canc. Res. , 58: 5271-5274 [1998]). In fact, primary cultures of malignant human mammary epithelial cells (BEC) have made great strides (Band et al., Canc. Res., 50: 7351-7357 [1990]; Ethier et al., Canc. Res., 53: 627-635). [1993]; Dairkee et al., Canc. Res., 57: 1590-1596 [1995]; Res., 58: 1408-1416 [1998]).
[0131]
The in vivo environment associated with malignant cells is believed to be composed of regions of low oxygen (ie, low oxygen), low pH, low levels of glucose, and high levels of metabolic waste. Simulating these conditions in culture has shown that a relatively pure population of primary breast cancer cells can be isolated because non-malignant cells cannot survive in these harsh in vitro environmental conditions (Dairkie et al., Supra). . It is widely believed that malignant BCE grows faster than non-malignant cells, which is a misconception. Therefore, another advantage of providing a culture of malignant cells in a harsh environment is that there is no non-malignant epithelial cells with a fast growth rate and they cannot grow beyond tumor cells. Also, drug resistance of tumor cells may be dependent on severe microenvironmental stress and may be the result of severe microenvironmental stress (Tomida and Tsuruo, supra). For this reason, it was anticipated that accurate measurement of drug resistance would require that cells be assayed in a suitable microenvironment. In fact, some of the most successful in vitro chemosensitivity tests used cells cultured in small capillaries sealed at both ends and incubated for 14 days. The resulting microenvironment in the capillaries is likely to be even more severe than that typically obtained with the sandwiched coverslip method (see Dairkee et al., Supra).
[0132]
Furthermore, in culture, the cells obtained (up to 10⁷) Are very similar to the original tumor and show one or more tumor phenotypes, including growth on soft agar. In fact, there is a suboptimal nutrient depletion environment in breast cancer (Dairkie et al., Supra, Tomida and Tsuruo, supra, and Brown and Giaccia, supra). The present invention provides a means to enhance the culture of primary breast cancer specimens. For example, tumor cells and normal cells are₂As a function of level, it is expected to exhibit very different pH behavior. In addition, it is highly likely that the pH range in which each cell tumor shows resistance is different. The net effect is that in a single incubator (in separate flasks, in the media preferred by each cell tumor), a common CO₂The level will not exist. In fact, the pH range over which normal cells are resistant is expected to be much narrower than that of cancer cells. No normal cell line has been found that favors a low pH environment (<7.1).
[0133]
Thus, if it is true that successful growth of human breast cancer epithelial cells requires a pH <7.0, it is also clear why previous in vitro chemosensitivity assays were not successful. Most media and cell incubation conditions have been designed to have a pH> 7.0. Thus, when human tumor biopsy specimens are cultured under these conditions, it is normal cells, not cancer cells, that attach and proliferate. As a result, one would unknowingly perform a chemosensitivity test on normal cells instead of cancer cells. To date, this may be the reason these assays have failed in clinical trials. However, use of the present invention does not require an understanding of the mechanism, and the present invention is not limited to any particular mechanism. However, the present invention monitors the morphological transformation of normal and abnormal cells as a function of pH and other microenvironmental aspects (eg, via a VSOM system capable of segmenting cells in transmitted light), Provides a means for analysis and quantification.
[0134]
K. Repetitive observation and analysis of specific fields of cells
The invention further provides a means for repeatedly observing and imaging multiple fields of cells in a closed tissue culture flask in a manner similar to that described herein for the multi-well plate. Having the ability to segment cells in transmitted light allows the VSOM system to return to the same field of view of the cells when the flask is returned to the microscope stage, acquire new images and quantify morphological changes As such, it means that the shape, position, and relative position of the cells can be preserved. This is an example of a more mechanical application of visual servoing, where the system relocates the same field of view of the cell even when the number, relative position, and shape of the cell change.
[0135]
Thus, in some embodiments of the present invention, the system provides a means of positioning relative to a very fine grid ("+"), lightly etched into a sterile culture dish using a fine finishing wheel for slicing. I will provide a. The culture dish may then be removed and returned to the incubator or stored under appropriate conditions. When re-imaging cells, the system is calibrated with respect to the "+" on the culture dish and the coordinates (0,0) are set to the center of the "+". The user can easily see the etching under transmitted light for initial alignment. The schematic diagram of FIG. 52 details the use of an etched culture dish to establish a frame that serves as a reference for the location of cells in the culture dish. In this way, the goal is achieved of annotating the response of each cell at its physical location in the culture dish. In addition, etching makes it easier to observe the same visual field of the cell several hours or several days later to monitor the response of the cell to a stimulus or the like over time.
[0136]
The system used in these experiments refers to its position relative to a very fine grid ("+"), lightly etched into a sterile culture dish using a fine finishing wheel for slicing. . The culture dish may then be removed and returned to the incubator or stored under appropriate conditions. When re-imaging cells, the system is calibrated with respect to the "+" on the culture dish and the coordinates (0,0) are set to the center of the "+". The user can easily see the etching under transmitted light for initial alignment. The schematic diagram of FIG. 52 details the use of an etched culture dish to establish a frame that serves as a reference for the location of cells in the culture dish. In this way, the goal is achieved of annotating the response of each cell at its physical location in the culture dish. In addition, etching makes it easier to observe the same visual field of the cell several hours or several days later to monitor the response of the cell to a stimulus or the like over time.
[0137]
Accordingly, the present invention provides a method of identifying and exploiting differences between cell types to favor a particular cell (ie, primary breast cancer cell) while favoring other cells (ie, normal mammary epithelial cells) for growth. And a VSOM fluorescence assay that can quantify the growth rate of living cells on a cell-by-cell basis.
[0138]
L. Enhanced in vitro drug and chemosensitivity testing
The present invention provides in vitro drug testing that, in addition to improving the means of proliferation and analysis of malignant cells, facilitates patient-oriented therapies to improve the clinical outcome of cancer patients (eg, breast cancer patients). And also greatly enhances chemosensitivity assays.
[0139]
A summary of the clinical correlations can be made using the following nomenclature: TP (true positives, patients whose cells are sensitive and respond to chemotherapy in vitro), TN (true negatives, cells resistant in vitro) FP (false positive, cells whose cells are resistant in vitro but clinically resistant), FN (false negative, cells are resistant in vitro but clinically resistant) Is the responding patient), PPA (positive prediction accuracy) = TP / (TP-FP), the percentage of patients who are sensitive in the test system and responded to the treatment, and NPA (negative prediction accuracy) = TN / (TN + FN ), Percentage of patients who were sensitive in the test system and did not respond to treatment.
[0140]
Historical correlation data between in vitro test results and patient response has shown an overall PPA of 72% and NPA of 90%. This corresponds to a sensitivity (ie, the ability to detect clinically responsive patients) of 85% and a specificity (ie, the ability to detect non-responsive patients) of 89%. These numbers are the average of seven assays in a study of 4263 patients, and the results are pooled from multiple individual studies (DeVita, "Cancer: Oncology Principles and Practices ( Cancer: Principles and Practice of Oncology), Lippincott-Raven, Philadelphia [1997]). The authors refer to these assays as "drug-responsive" rather than "chemotherapy" assays because they are better at predicting patient drug resistance than patient response. Has been advocated.
[0141]
Digital imaging fluorescence microscopy, real-time analysis of digital images, and a bioinformatics database, combined as robotic vision technology “visual servo control (VS)”, provide dynamics of experimental parameters based on analysis of digital image content. Operation. In addition, VSOM has the ability to scan back and forth through multiple fields of view, and thereby to monitor a very large population of living cells.
[0142]
As described above, VSOM further provides the ability to repeatedly return to the same field of the cell of interest to monitor changes on a cell-by-cell basis. This ability facilitates correlating the early physiological response observed in the cell with the future biological state or terminal fate of the cell, whether it is proliferation or apoptosis or arrest at some cell cycle stage. In fact, one aspect of the present invention provides the most desirable predictive assays. Importantly, these assays are dynamic and adaptive.
[0143]
In one embodiment, an adaptive VSOM fluorescence assay observes an early physiological response of cells to intelligently selected computer-controlled stimuli. In some preferred embodiments, these stimuli and perturbations are reversible and do not permanently subject the cells to a particular biological pathway. For example, a particular subpopulation of cells is identified based on the observed stress response while reducing the pH of the extracellular medium. One of the intelligent reactions of the system could be to lower the pH of the medium until the cells of interest are identified, and then return the cells to a higher pH before irreversible damage to the cells occurs. The cells of interest are then identified on a cell-by-cell basis, and the system is used to perform additional tests to optimize and / or utilize this differential response to low pH environments.
[0144]
The present invention also provides the ability to repeatedly probe cells to search for physiological features available through anti-cancer agents or drug combinations. Thus, the VSOM fluorescence assay of the present invention, in a preferred embodiment, is a means of rapidly identifying cells using a complex mixture of cells, wherein the fluorescent probe is non-toxic to the cells and the applied stimulus or Cells are identified based on the characteristic physiological response to the perturbation and provide a means by which the stimulus or perturbation has no irreversible effect on the cells. In fact, in a particularly preferred embodiment, the predictive assay of the present invention is a dynamic and adaptive VSOM fluorescence assay that involves observing a series of early physiological responses to intelligently selected computer-controlled stimuli. This approach is even more powerful if such stimuli and perturbations are reversible and do not permanently subject the cell to a particular biological pathway. The ability to repeatedly apply relatively harmless stimuli and perturbations to living cells allows the cells to be repeatedly probed for the physiological features available with anti-cancer agents or combinations of anti-cancer agents. The VSOM fluorescence assay of the present invention can better predict the future behavior and / or terminal fate of a cell. For example, the invention provides a means of determining whether a cell has been stimulated sufficiently to proliferate or has been stressed to the point where apoptosis is necessary. This ability depends on the early physiological responses observed in the cell and the future biological state or ultimate biological fate of the cell (ie, proliferation, apoptosis, arrest at a particular cell cycle stage, etc.). Helps to correlate.
[0145]
The present invention further provides an object-oriented database that contains a record of previously observed physiological cellular responses. This database provides a very useful resource in performing knowledge-based control of VSOM assays. Having access to this database remotely (i.e., via any VS-enabled microscope) allows for control of these devices over the Internet and the necessary information to operate these devices and them. The functionality of these devices can be greatly enhanced while significantly reducing the cost of computer performance.
[0146]
At least three prospective clinical trials have attempted to use in vitro chemosensitivity tests to improve the response and survival of cancer patients to chemotherapy. However, these studies did not show any improvement in the survival of patients receiving chemotherapy based on the results of the in vitro chemosensitivity test. However, each patient is at least slightly different from any other patient. For example, one patient may have an adverse reaction to a drug and another patient may not experience any side effects. If the patient has an adverse response to the drug or if the patient's tumor does not respond to the drug, the physician will change the drug and / or treatment regimen to identify an effective drug. In the case of breast cancer, the physician chooses a drug (or two or more drugs) from a wide range of possibilities and chooses the combination that is most effective against the individual patient's tumor and has minimal side effects in the individual patient. I have to find it. Often, physicians are forced to try a series of drugs for each patient (ie, these are in vivo drug response tests to determine the optimal treatment regimen). In vitro drug response testing has the advantage that a wider range of drugs can be tested in a shorter time without the patient undergoing a series of debilitating trial and error procedures to find a suitable treatment. Thus, for over 40 years, attempts have been made to remove tumor cells from a patient's body and test them in vitro.
[0147]
The two in vitro chemosensitivity assays used in these unsuccessful methods are HTCA (human tumor cloning assay) and DiSC (spread cytotoxicity assay). The techniques of in vitro cell culture and proliferation used in these assays, as well as the biological endpoints measured, vary significantly. HTCA requires a culture in which minced tumor tissue is placed in very small capillaries and sealed at both ends and incubated for 14 days. This mixture containing malignant and non-malignant cells (cell number 0.2-1.0 × 10⁵Before sealing the cells in a capillary, the cells are exposed for 1 hour to a standard anti-cancer drug adjusted to a concentration corresponding to one-tenth of the peak concentration observed in human plasma. The cells are then suspended in 0.3% agar and sealed at 37 ° C. and 7% CO 2 in a 100 μl capillary.₂For 14 days. After 14 days, cells are extracted from the capillaries and the number of colonies is determined manually using a microscope. This capillary cloning system was used because it required a smaller number of tumor cells initially and improved the generation of colonies with significant numbers of cells.
[0148]
The DiSC assay is a non-cloning assay in which tumor cells are cultured in liquid media in small polypropylene tubes for 4-6 days to amplify the number of cells available for testing. Exposure time to the drug varies from drug to drug and ranges from 1 hour to 4 days. Generally, drug concentrations are determined empirically and are higher than those used in HTCA. For example, cells are treated for 1 hour with 0.04 μg / ml doxorubicin for HTCA, whereas 1 hour for 1.2 μ / ml for DiSC. In the DiSC assay, cell membrane integrity is assessed by staining dead cells in a suspension containing Fast Green. Acetaldehyde-fixed duck red blood cells (DRBC) are added to the culture as an internal standard, and the entire mixture is cytocentrifuged. Live cells are colorless, dead cells and DRBC are stained green on microscope slides. Will be collected. The slides are then counterstained with HE (Hematoxylin-Eosin) to stain live cells. Next, a skilled technician microscopically identifies the cells as tumor cells or normal cells. Next, the ratio of living cells to DRBC is determined.
[0149]
A common group of problems has been pointed out for the data obtained in the above tests. Both assays are manual and labor intensive, rely on human judgment for quantification, and often do not yield results due to poor growth of some patients' tumor cells in these in vitro systems Sometimes. For these reasons, clinicians have low awareness and acceptance of these assays. This makes it difficult to collect patients and specimens. However, these tests are replacing "second generation" tests that are better at identifying drugs that have no effect on patient tumor cells. By excluding some drugs in this way, physicians do not need to take these drugs into account when conducting in vivo drug response tests in patients. These tests are based on technology that is labor intensive but has been available for many years. However, these tests are labor intensive and have the potential for human error.
[0150]
It is possible that such studies to date have not resulted in a dramatic improvement in patient response and / or survival as a result of various factors, including microenvironment such as tumor The cell's preferred microenvironment was not properly simulated in vitro), the physiological response of individual cells was not closely monitored during the assay, and / or drug concentrations, combinations, and exposure times were limited And was not scientifically optimized.
[0151]
However, during the development of the present invention (ie, one of the "third generation" tests), advanced cell culture techniques and innovative technologies from the field of robotic vision and digital imaging fluoroscopy have been combined. Was considered. The visual servo-controlled optical microscope (VSOM) of the present invention is useful, for example, for the rapid discovery of important physiological features that distinguish between malignant and non-malignant cells (eg, breast epithelial cells (BECs)). As described above, VSOM is based on the rapid and automatic generation of an in vitro microenvironment that favors the growth of cells of interest (eg, tumor cells), and on the basis of early physiological responses to applied stress. To generate in vitro fluorescence assays that predict cell behavior. Accordingly, the present invention provides a means of overcoming the problems associated with the aforementioned clinical trials, including the difficulties in growing tumor cells in vitro, Includes limited ability to observe cells, limited number of biological endpoints that can be observed in a single assay, and lack of automation. The present invention also provides a means for rapidly finding and optimizing microenvironmental conditions suitable for in vitro growth and chemosensitivity testing of primary breast cancer cells (eg, human cells). Further, the present invention provides a means for achieving this purpose on a tumor-by-tumor basis.
[0152]
M. Monitoring of microenvironment parameters
The present invention provides a means to use biosensors and related electronics necessary for online monitoring of important microenvironmental parameters useful for VSOM experiments. The sensor used in these experiments is expected to be a ZABS flow-through sensor suitable for serial connection to a perfusion line. Therefore, it is possible to continuously monitor and record several important properties of the medium just before reaching the cells. In some embodiments of the present invention, the presence of large transmitted light collectors, temperature probes, and vacuum suction devices provides sufficient space to place all of these sensors in the actual cell perfusion chamber. Absent. Therefore, while continuously perfusing the medium into the microincubation chamber, the pH of the medium, L-lactic acid or glucose, extracellular calcium, and pO₂Monitor
[0153]
In some embodiments, the VSOM method uses four computer-controlled syringe perfusion pumps. In addition, the microincubation chamber was modified to fit securely on a computer controlled XY scanning stage. A vacuum driven suction device eliminates excess liquid and a temperature probe provides feedback for temperature control.
[0154]
In various aspects, an electronics package is used with the present invention. For example, the electronics package “PLUGSYS” is used in these experiments. PLUGSYS is a modular design whose basic system cases are pH, L-lactate or glucose, extracellular calcium, and pO₂Etc. have enough slots to hold the various amplification modules needed for measurements. The PLUGSYS case is also compatible with the data acquisition hardware used, such as a PC PCI
An A / D conversion board is included. This device provides important real-time data for VSOM system control and is required for online monitoring of microenvironment parameters important for conducting and interpreting VSOM experiments and important microenvironment parameters used during VSOM experiments. Enables continuous monitoring and recording of relevant electronic devices.
[0155]
Experiment
The following examples are provided for the purpose of illustrating and further illustrating certain preferred embodiments and aspects of the present invention, and thus should not be construed as limiting the scope of the invention.
[0156]
The following abbreviations are used in the following experimental disclosures: VS (visual servo control); VSOM (visual servo control optical microscope / microscopy); ° C. (degrees Celsius); rpm (revolutions per minute); BSA (bovine serum albumin); H₂O (water); aa (amino acid); bp (base pair); kb (kilo base pair); kD (kilodalton); gm (gram); μg (microgram); mg (milligram); μl (microliter); ml (milliliter); mm (millimeter); nm (nanometer); μm (micrometer); M (mole); mM (millimole); μM (micromole); (Volts); MW (molecular weight); s and sec (seconds); min (s) (minutes); hr (s) (hours);₂(Magnesium chloride); NaCl (sodium chloride); OD₂₈₀(Optical density at 280 nm); OD₆₀₀(Optical density at 600 nm); [Ca⁺²]_in(Intracellular calcium concentration); [Ca⁺²]_ex(Extracellular calcium concentration); DOX (doxorubicin); BEC (breast epithelial cell); HMEC (human mammary epithelial cell); MCF-7 (one of human breast cancer cell lines); MCF-7WTC (MCF-7 wild type, Cowan); HTCA (human tumor cloning assay); DiSC (staining cytotoxicity); CAM (calcein-AM); DRBC (duck erythrocytes); DS (drug sensitivity); MDR (multidrug resistance); MEBM (mammary epithelial cell growth medium); MEGM (mammary epithelial cell growth medium); PBS (phosphate buffered saline); D_PBS (Dulbecco's PBS); FN (false negative); TN (true negative) FP (false positive); TP (true positive); NPA (negative prediction accuracy); PPA (positive prediction accuracy); PMA (phorbol 12-myristate 1) TRME (tetramethylrhodamine ethyl ester); TG (thapsigargin); FURA-2-PE3 (one of the calcium-sensitive fluorescent dyes); H42 (Hoechst 33342); LT-R (Lysotracker-). Red); FNA (fine needle aspiration); PAGE (polyacrylamide gel electrophoresis); PBS (phosphate buffered saline [150 mM NaCl, 10 mM sodium phosphate buffer, pH 7.2]); PCR (polymerase) Chain reaction); PEG (polyethylene glycol); SDS (sodium dodecyl sulfate); Tris (tris (hydroxymethyl) aminomethane); DMSO (dimethylsulfoxide); w / v (mass-volume ratio); v / v (volume-volume ratio) ); Amersham (Amers am
Life Science, Inc. ICN (ICN Pharmaceuticals, Inc., Costa, CA); Arlington Heights, Illinois;
Mesa); ATCC (American Type Culture Collection, Rockville, MD); Becton Dickinson (Becton)
Dickinson Labware, Lincoln Park, NJ; BioRad (BioRad, Richmond, CA); Clonetics (Clonetics, San).
Diego); Clontech (CLONTECH Laboratories, Palo Alto, CA); Molecular Probes (Molecular)
Probes, Eugene, OR); GIBCO BRL or Gibco BRL (Life Technologies, Inc., Gaithersburg, MD); Invitrogen (Invitrogen).
Corp .. New England Biolabs (New England Biolabs, Inc., Beverly, Mass.); Novagen (Novagen, Inc., Madison, Wisconsin; Madison, Wis.);
Chemical Co. St. Missouri. Louis); Stratagene (Stratagene Cloning Systems, La, CA).
Jolla); Creative Scientific Methods (Creative Scientific Methods, Inc., www.cre8ive-sci.com); and Zeiss (Carl).
Zeiss, Inc. , Thornwood, NY).
[0157]
Example 1
Imaging protocol to repeatedly return to the same cell set
These experiments describe methods for repeatedly returning to the same set of cells in culture. By using these methods, one or more sets of live cells in a cell culture dish placed on a moving microscope stage can be observed. Importantly, this allows the culture dish to be removed from the stage and placed in an incubator for purposes such as growing cells. The culture dish can then be placed on a microscope stage and the same set of cells can be automatically subjected to observation. These steps can be repeated as desired or required.
[0158]
In these experiments, the reaction of a cell set (500-1000 cells at 10X magnification) can be monitored while various compounds are perfused into a culture dish. A stack of images consisting of a specified number of channels is acquired one or more times at specified intervals over a specified length of time. The editable (text) recipe file specifies the experimental recipe to be used. A stack of only one image is acquired for a "snapshot", and a complete stack of images can be acquired at regular intervals for a "timelapse" experiment. Because it is preferred that the user does not move the stage or culture dish between experiments in a continuous time-lapse experiment, the present invention provides for multiple time-lapses of the same set of cells in succession before moving the culture dish. Makes it easy to conduct photography experiments. After performing the desired analysis, the culture dish is removed from the microscope stage and placed in an incubator for hours or days as needed. When the culture dish is returned to the microscope stage, the cells may be alive or dead (eg, fixed or stained cells). Of course, no time-lapse photography experiments are performed when dead cells are used. However, in these experiments, the pericellular field of view also acquires a “snapshot” image stack to ensure that a single cell set that is monitored repeatedly is representative of many surrounding cell sets It is imaged by doing.
[0159]
In these experiments, the "channel indicator" is specified by _0, _1, _2, _3, _4, _5, and _X in the file name. These numbers indicate the particular location of the optical filter wheel, and _X indicates the transmitted light image. The identification of the optical filter at each position of the filter wheel is described in the header of each image according to the wavelength. The type of transmitted light used is not specified. For example, "dc122999dc2.00001_0ics" indicates that an image has been acquired at the position "0" of the filter wheel. Based on the image header, the filter at this location is a 360 nm fluorescence excitation filter. The channel indicator “dc122999dc2.00001_X.ics” indicates that the image was acquired with the transmitted light shutter opened. This image may be a phase contrast image, DIC, bright field image, or the like. The identification of the maximum wavelength of the optical filter at each location of the filter wheel used in these experiments is described in an editable config text file, described in more detail below.
[0160]
"Cell set" refers to all cells that are visible within a particular field of view in a digital image at a particular magnification. Over the course of hours and days, a single living cell may grow into and divide into two, degrade cells, die and float, and move individual cells. The initial position of the cell set is specified as relative to the (0,0) point, which is the "+" (cross-hair) symbol physically etched into the culture dish (eg, Petri dish) surface ( See FIG. 52).
[0161]
An "image stack" is a set of all specified channels at a single point in time. For example, a seventh stack in a time-lapse series of image stacks may be as follows:

[0162]
The “recipe file” is a text file that specifies a protocol to be used. For example, a recipe file contains:
Which fluorescent channels are included in the complete image stack;
Exposure time for each channel;
The order in which the image channels are acquired;
Interval between image stacks acquired in time-lapse experiments;
The number of stacks to get;
Identification of the excitation filter at each position of the filter wheel.
[0163]
Multiple iterations are possible during the implementation of the method. For example, "first iteration" uses a culture dish of cells that have not been imaged so far. The (0,0) mark is set on the computer screen using the crosshairs etched on the culture dish containing the cells. The user looks for a set of cells to observe. At this stage, various shutters are opened and closed, the filter wheel is moved to various positions, and the program must keep track of where the user has moved the culture dish relative to the (0,0) mark. . The user then focuses on the set of cells of interest. Next, an experiment is performed based on the information of the recipe file. This initial experiment is typically performed as a live-cell time-lapse assay.
[0164]
However, in other embodiments, a series of "snapshots" of fixed or stained cells is used. The recipe file is then re-edited without moving the cell or stage (refocusing is allowed) and one or more time-lapse experiments are performed. The culture dish containing the cells is then removed and placed in an incubator (or other storage location) until needed. In some embodiments using "snapshots", the user moves the culture dish to view another set of cells and an additional snapshot is taken.
[0165]
In the "second iteration", the cell set is observed again. This cell set has already been imaged and information about its location has been stored. First, a culture dish containing cells to be analyzed is placed on a microscope stage. The first transmitted light image of the crosshair mark is acquired. The user "clicks" on the (0,0) mark on the screen and tells the program which cell set to observe. The program moves the field of view to the (0,0) mark and then to the appropriate cell field of view. The user focuses on the cells and performs an experiment based on the information in the edited recipe file. At this stage, the user opens the shutter and moves the filter wheel to focus. The recipe file is re-edited without moving the cell or stage (refocusing is allowed) and one or more additional time-lapse experiments are performed. In a snapshot experiment, the field of view is changed as desired, and a desired number of additional snapshots are taken. At the end of the experimental protocol, the cell culture dish is removed from the stage and returned to a storage location (eg, incubator). For the "nth iteration", a set of live cells is repeatedly imaged and information about cell location is stored. The sequence of events is the same as for the second iteration.
[0166]
Example 2
Use of VSOM
In this embodiment, a VSOM device and other aspects of the present invention will be described.
[0167]
A. VSOM system optical platform
In most embodiments, the VSOM system focuses on inverted fluorescent microscopes because, in general, it is easier to design the cell chamber and cell container when the microscope is in an inverted shape (ie, the objective lens faces upward). Be composed. However, there are also cell chambers that can be used with an upright microscope (a microscope with the objective lens pointing down). The VSOM system of the present invention is suitable for use with such an optical platform. Preferred is a fluorescence microscope of a grade suitable for biological research that has sufficient weight to be stable when the camera and microscope peripherals are mounted on the microscope. A standard 10 × objective is sufficient for some VSOM experiments. Research grade microscopes, objectives, and various peripherals are available from Carl Zeiss.
Zeiss, Inc. Manufactured by a variety of manufacturers, including Thornwood, NY, and Nikon USA, Melville, NY. Environmentally controlled cell chambers are available from Zeiss, Harvard Apparatus.
Apparatus, Holliston, MA) and Biotechs, Inc. (Biological Optical).
It is manufactured by a variety of different manufacturers, including Technologies (Butler, PA).
[0168]
B. VSOM peripheral device
Digital cameras that can be used in a preferred embodiment of the present invention include those designed for low-light fluorescence microscopy applications (eg, Roper Scientific).
Scientific Inc. (Tucson, Ariz.) And Quantitative Imaging Corp. (Burnby, Canada). Such a camera has a science-grade CCD with 12-bit digitization capability and has a total of about one million pixels, which is about 7 μm × 7 μm in size. Such cameras include software development kits, including C ++ source code and software development kits, or software that integrates camera operation with software, optical, electrical, and mechanical elements (ie, microscope peripherals) on a host PC. Available with a host connection kit that can be used to describe. These cameras are provided with a variety of industry standard computer interfaces, such as a PCI computer interface card or a directly connected "FireWire" (IEEE 1394) cable connection, so that the functions of the camera can be programmed via a host computer.
[0169]
In addition to digital cameras, other peripherals can be used with the present invention, including XY scanning stages, optical filter wheels, coarse Z-axis focus motors, and transmitted light and surface fluorescent shutters. (For example, LUDL Electronics Products (LUDL
Electronics Products, Ltd. Peripherals available from manufacturers such as Hawthorne, NY). Various manufacturers provide other equipment. For example, a shutter instrument (Sutter)
Instrument Co. Novato, Calif. Manufactures optical filter wheels and robotic micromanipulators, and syringe pumps are manufactured by Harvard Apparatus.
From Apparatus, Holliston, Mass., The microscope objective lens nanopositioner (piezoelectric positioner) is a Physik instrument.
Intrumente, Waldbron, Germany). These companies also sell the appropriate programmable controller for each peripheral device. Harvard Apparatus also sells PLUGSYS and similar measurement and control devices for operating and interfacing various environmental sensors (oxygen, lactate, glucose, pH, etc.) with a host computer.
[0170]
C. Biological reagent
Biological reagents and fluorescent labels and the like are available from a number of manufacturers. For example, fluorescent probes for live cells are available from Molecular Probes (Eugene, OR). Other cell culture media, equipment and supplies for cell growth, and related equipment and supplies are available from standard suppliers (eg, Sigma, Fisher, etc.).
[0171]
D. VSOM host computer
In a preferred aspect of the invention, one or more available PCI slots (depending on whether the digital camera requires a PCI slot to interface with a host computer), standard serial and parallel ports, OHCI compatible IEEE 1394 port (if a FireWire enabled digital camera is required), 256 MB of RAM, one or more 18 Gbyte hard drives, tape backup unit, Ethernet adapter, 32-bit color video card , A high resolution monitor (preferably 1280 x 1024 pixels), and a version of LINUX or Microsoft
Requires a 450 MHz Pentium III class industry standard PC running Windows. Standard C ++ computer programming techniques are required to integrate the operation of digital cameras and various microscope peripherals with the aforementioned software and algorithms for online image segmentation and analysis. Therefore, standard C ++ programming software, and Microsoft
An integrated software development package such as Visual Studio C ++ is required. In addition, a software package that includes a C ++ library for standard digital image acquisition, peripheral control and automation, and standard image processing, plotting, and display operations is desirable. As such a package, "SCIL
Image "(TNO, Delft, The Netherlands)," MetaFluor "and" Metamorph "(Universal Imaging Corporation).
Imaging Corporation, Downingtown, PA) and "Component Works ++" (National Instruments, Inc.)
Instruments, Austin, Tex.). Further, in some cases, as the number of microscope peripherals and the number of environmental sensors controlled by the VSOM system increase, standard analog-to-digital interface cards such as those manufactured by National Instruments are required. Become.
[0172]
E. FIG. Implementation of VSOM
In addition to the optical platform, minimal implementation of VSOM requires at least one of each of the following: a temperature-controlled cell chamber, a computer-controlled syringe pump, an illumination shutter, and an optical filter wheel. Minimal experiments do not require the ability to handle transmitted light, only fluorescent microscopy. The only microenvironment controller needed is a temperature control unit capable of maintaining the temperature at 37 ° C. with a temperature probe that can be inserted into the specimen container. In experiments performed at a minimum, no Z-axis adjustment of the objective lens is necessary, assuming that the optical platform is relatively stable and free of vibration. Since only a single cell field (not multiple fields) needs to be observed, no XY positioning stage is required. However, the invention is not limited to this particular design type.
[0173]
In one embodiment, the cells are grown in a standard cell incubator in a standard plastic petri dish using standard techniques prior to the experiment. In some embodiments, the cells are labeled prior to the experiment with a nuclear staining method suitable for living cells. For example, Hoechst 33342 (H42) is used to stain the nuclei of living cells in the following manner. Cells that are allowed to attach to an appropriately sized specimen container (eg, a round 35 mm cell culture dish) for 24 hours are exposed to 1.0 μg / mL H42 for about 90 minutes in a cell incubator. After this exposure, the phenol red indicator present in the H42 and cell growth medium is washed away and replaced with fresh medium without phenol red or excess fetal bovine serum. The specimen container is placed in the cell chamber, the operator focuses on the appropriate cell field, and then sends the image output to a digital camera. One of the computer-controlled syringes is filled with an appropriate solution that does not contain a fluorescent probe (serum or phenol red-free, cell culture medium, etc.), and the other syringe contains the same solution and a fluorescent probe that is suitable for live cells. Is filled. For example, calcein-AM (CAM) is a suitable probe that can indicate the presence of multidrug resistance (MDR) in a cell line. In a minimal VSOM-enabled computer controlled MDR assay, about 1.0 μM CAM solution is perfused into the cell chamber at a flow rate of about 0.5 mL per minute. The compound crosses the membrane of a living cell and is then metabolized, becomes fluorescent, and is retained inside the cell. As is well known to those skilled in the art, if a cell has one or more MDR proteins, the cell has the ability to export CAM out of the cell. In these cases, the rate at which the cells take up the compound in the presence of the CAM is reduced, and the cell retains less CAM after the CAM has been flushed from the chamber.
[0174]
The VSOM system initiates perfusion of the CAM solution and uses the intranuclear H42 blue nuclear fluorescence signal to continuously segment the live cell field of view. Thus, a contour describing the nucleus is acquired in the blue channel. These same contours are also used in the green channel to calculate the average fluorescence intensity of the CAM in the nuclear region. Since CAMs are often uniformly distributed throughout the cell, it is not necessary to detect cytoplasmic regions in addition to nuclear regions. Images are acquired at appropriate intervals (eg, every 60 seconds), and thus the average intracellular fluorescence intensity (MI) in the green channel is observed for the nuclear region of each cell.
[0175]
Next, the computer algorithm performs the following servo loop operation to adjust the amount of CAM in the cell chamber in the following manner. The average MI of all cells (as defined above) is monitored for each image acquisition point, and then the following is performed:
a) when the average MI of all cells in the field of view reaches a user-defined threshold, the syringe with CAM is shut off and the syringe without CAM is turned on; or
b) the syringe without CAM is shut off when the average MI of all cells in the field of view is reduced by a user-defined percentage from the maximum average MI reached in (a);
c) End of experiment.
[0176]
In such a minimal experiment, individual rates of CAM uptake and retention are obtained for each cell. The system can then report the number (or percentage) of cells that showed MDR in the field of view at the end of the experiment. In addition, such cells are thereby identified and can be subjected to subsequent tests to test for compounds that alter MDR. However, the invention is not limited to this particular system, format, cell, etc. In fact, the present invention offers the user maximum flexibility in terms of experimental design.
[0177]
Example 3
VSOM experiment
In this embodiment, a VSOM experiment performed by adding the following changes to the description of the second embodiment will be described.
[0178]
The optical platform used is a Zeiss Axiovert 135 H / DIC, a TV inverted microscope equipped for transmission light microscopy (phase contrast and DIC) and multicolor fluorescence microscopy. The microscope includes a computer controlled XY scanning stage, a Z-axis stepper motor, and a 6 position filter wheel (LUDL Electronic Products (LUDL)
Electronic Products, Ltd. , Hawthorne, NY)). In these studies, a 12-bit Xilix with a Kodak KAF-1400 CCD chip (1317 × 1035 pixels, 7 × 7 micron pixel size) was used.
A CCD camera (Xillix Technologies, Vancouver, British Columbia) was used. This camera has a readout speed of 8 MHz (ie, about 4 full-size images per set). The host computer is a multitasking UNIX workstation, Sparkstation.
Ultra 1 was set, and the image output from the computer camera was directly read by the host computer.
[0179]
In addition, a Peltier temperature controlled microperfusion chamber (PDMI-2 open chamber with TC-202 bipolar temperature controller, Harvard Apparatus / Medical Systems Research Products).
Apparatus / Medical Systems Research Products, Holliston, Mass.) And dual syringe pumps (Pump-33, Harvard Apparatus) were used. The syringes need not be the same size, and each syringe has its own computer-controlled block so that the perfusion rate can be controlled independently. A "bath" type thermistor (BSC-T3, 36K ohm total) was used with PDMI-2.
[0180]
In some preferred embodiments, the inverted design of the microscope allowed perfusion of live cells, and immobilized cells were imaged from below. In some aspects of the invention, an oil immersion objective can be used. Also, cells are grown in 35 mm Petri dishes in some embodiments and on circular cover slips in other embodiments and viewed from below. With the configuration of this inverted microscope, the user can switch from fluorescence to transmitted light detection (phase difference or DIC), and can confirm the position and morphology of cells. This is achieved by a computer controlled transmitted light shutter. A computer controlled surface fluorescent shutter on the filter wheel can also be used (LUDL). In these experiments, the automatic visual servo control operation of the system (ie, control of the pump, illumination shutter, filter wheel, and Xilix camera) was performed with the program "ADR". In a VSOM experiment performed using a pre-programmed recipe, the program "VX5" was used.
[0181]
A. ADR visual servo control experiment
MCF-7 WTC cells were grown in 35 mm cell culture dishes and prestained with H42. After preincubation with H42 for about 1 hour, the cells were rinsed with DPBSGP and placed in a warm microscope chamber. The ADR program worked without any problems as follows. During this experiment, three digital images (360 nm excitation, 490 nm excitation, and transmitted light, bright field) were acquired at specified regular intervals and all stained with H42 before the next imaging interval. Nuclei were detected and segmented. Using the outline of each nucleus depicted in the blue channel, the value of the mean cell fluorescence intensity was calculated from the green channel (MI). An initial perfusion with the DBPSGP solution alone (syringe # 1) was performed at 1.0 mL / min for 8.3 minutes (as specified by the user), and then syringe # 1 was stopped. Next, perfusion of a DBPSGP solution (syringe # 2) containing 1 μM CAM was started at 0.5 mL / min. When the average MI of all cells reached the specified threshold, syringe # 2 was stopped by normal visual servo control. 12. By the time cells have accumulated sufficient CAM and the system has detected sufficient intracellular CAM (in the nuclear region) and the system has successfully stopped syringe # 2 and restarted syringe # 1. It took 9 minutes. After 11.4 minutes (as specified by the user), the system again stopped syringe # 1. The system continued to monitor the cells in the chamber (as specified by the user) and another 24.9 minutes elapsed without flow. Next, at the end of the experiment, the system stopped acquiring images.
[0182]
Example 4
MDR Assay-Test Kit and VSOM Experiment
As described herein, the MDR assay beta test kit was also tested in the VSOM experiment. FIG. 11 shows a schematic diagram of this VSOM experiment. Each pump turns on and off at a specified flow rate according to a pre-programmed recipe or based on real-time analysis of individual cell responses. The concentrations of CAM, V and MK used are shown in the figure. The three exposures to the CAM were each 900 seconds (20 minutes), so the total exposure time to the CAM was 60 minutes. The black window at the bottom of the figure shows the average of the response of all cells in the field of view. This average response plot is displayed on the computer screen during the VSOM experiment, and a digital image of one or more channels is also displayed on the screen (typically, one of the displayed channels is transmitted light).
[0183]
During the experiment, the plots and images are continually redrawn and updated. In two plots recorded for MCF-7 (DS, run # 1) and MCF-7 ADR (MDR, run # 2), the amount of calcein accumulation in DS cells (240 ± 82, N = 336) Was found to be greater than the accumulated amount of MDR cells (103 ± 80, N = 161). This is consistent with the relative differences in calcein accumulation between DS and MDR cells observed in multi-well plate experiments representing the average of a large number of cells (see Table 1 above: 1658/4822 = 0. 34, 1965/6598 = 0.30, VSOM: 103/240 = 0.43). However, in the VSOM experiment, the average cell fluorescence of each cell was monitored, and deviation from the average value was easily detected. For example, several cells have begun to lose viability and plasma membrane integrity. Such cells cannot retain calcein, and calcein leaks, resulting in a reduced fluorescent signal. This may be an effect related to the toxicity of DMSO described above. However, an understanding of the mechanism is not required to use the present invention, and the present invention is not limited to a particular mechanism. Nevertheless, it should be noted that current VSOM systems have the ability to easily terminate exposure / perfusion once a single cell is found dead.
[0184]
FIG. 13 shows cell responses of four cells selected from the result of execution # 2 of the present example. These response curves are good examples of the principles underlying the beta test kit. In two of the response curves (FIG. 13), cells do not accumulate calcein until exposure to verapamil (V) during operation of syringe # 5 (FIG. 11). Since the Pgp pump is not inhibited by MK-571 (MK), it can be interpreted that these cells express only the Pgp pump. It is clear that none of these cells are present in the DS cell population (FIG. 12). In addition, after CAM + V exposure, one cell rapidly accumulated calcein until its mean cell fluorescence reached within one standard deviation of the mean cell response observed with DS cells (FIG. 12). Assuming all other factors are the same, the second cell can be expected to have a lower content of Pgp molecules. The remaining two curves (FIG. 13) show another response during exposure of CAM (only) and CAM + MK, which requires further analysis. The observed curves cannot be explained in terms of the presence or absence of only two pumps without the ability to excrete internalized calcein. However, an understanding of the mechanism is not required to use the present invention, and the present invention is not limited to a particular mechanism. This is a type of detailed physiological information not available in flow cytometry and multiwell plate assays. In contrast, the VSOM technology presented here has these capabilities.
[0185]
For example, detailed rate information regarding calcein accumulation and retention is available throughout the VSOM experiment. For some response curves, the percentage of calcein accumulation during three CAM perfusions can be calculated. In addition, calcein retention during BUFF perfusion can be calculated. Calcein accumulation was found to occur at 1500, 3500, and 5500 seconds, and plateaus corresponding to good calcein retention were observed at 2500, 4500, and 6500 second intervals. As mentioned above, some MDR pumps have the ability to excrete calcein after it has been internalized into cells. The ability of DS cells to excrete internalized calcein is significant during the 2500 second period corresponding to BUFF perfusion (see FIG. 12). However, two MDR cells (FIG. 13) do not have this ability.
[0186]
Further, using the above teachings, the pre-programmed manipulations in which the computer makes the decisions are suitable for visual servo control operations.
[0187]
Example 5
Digital imaging fluorescence method using modified reagent
VSOM can be used to automate a digital imaging fluorescence microscope. For example, calcein-AM (CAM, 0.25 μM) containing the modifying reagent MK-571 (10 μM) or verapamil (50 μM) can be perfused into the microperfusion chamber. MK-571 is a specific inhibitor of the transmembrane protein MRP, while verapamil inhibits the MRP as well as the transmembrane protein PgP. All of these transmembrane proteins excrete exogenous compounds, and these proteins are thought to be involved in multidrug resistance. Separate genes encode these two different proteins. The drug cross-resistance profiles (spectrum) of these proteins overlap but are not exactly the same. Furthermore, as mentioned above, these proteins have different sensitivities to various inhibitors.
[0188]
MCF-7ADR cells labeled with Hoechst 33342 were observed at 35 ° C., and the average fluorescence intensity (MI, calcein) for each cell was determined. The average fluorescence intensity per cell (MI, calcein) is calculated in real time, and the software operation of the syringe pump is performed based on the results of these calculations (ie, visual servo control). The protocol used in these experiments is based on the aforementioned MDR beta test kit (Molecular Probes). Images are acquired during three perfusion intervals: (1) 0-2100 seconds, CAM; (2) 8600-11,000 seconds, CAM + MK-751; and (3) 14,000-16,000 seconds. , CAM + verapamil. Typically, all cells will respond by the third perfusion interval. Subpopulations expressing MRP and / or PgP are inferred based on responses with MI> 200 as follows: (a) Response at 1 (neither PgP nor MK-571 expressed ), (B) does not react to 2 (MRP), and (c) does not react to either 1 or 2 (PgP only). During interval 1, neither expression of PgP nor MK-571 was observed, but in interval 2, expression of MRP was observed. Due to MRP inhibition by MK-571, the increase in MI due to calcein accumulation exceeded 30 in 30 cells. Three digital images (each channel was acquired every 60 seconds). For example, the MI of cell # 10 increased from 30 to 760 during this 2400 second perfusion interval.
[0189]
Example 6
BrdU proliferation assay for live cells
In this example, a BrdU proliferation assay system for living cells will be described. The ability to correlate a particular cellular response to a particular biological endpoint improves the predictive power of any VSOM prediction assay. Although there are multiple fluorescent assays for apoptosis, few fluorescent assays for live cells can measure important parameters such as BrdU incorporation that can be used to quantify DNA synthesis or cell proliferation. In these experiments, an image ratio calculation technique for detecting the growth state of living cells by quantifying the amount of BrdU incorporation was developed. The protocol for this assay is described in the published Dr. It was adapted based on the protocol of Paul Yaswen (Stampfer et al., Exp. Cell Res., 208: 175-188 [1993]). The same cell line and medium were used. However, for DNA synthesis assays, [³5-bromo-2'-deoxy-uridine (BrdU) was used instead of [H] -thymidine. These experiments were performed using a 5-bromo-2'-deoxy-uridine labeling and detection kit (No. 1296736, Boehringer Mannheim). After VSOM experiments, cells were fixed and VSOM results were confirmed based on digital image ratio calculations using a conventional indirect immunofluorescence assay.
[0190]
Cells stimulated to proliferate double their normal complement of DNA in preparation for cell division. These actively growing cells take up A, T, C, and G nucleotides from the extracellular medium and can use them for DNA synthesis. However, in the presence of BrdU, BdU is used for DNA synthesis instead of T. Thus, when a population of cells is exposed to BrdU for a short time (eg, 1 hour), most BrdU is incorporated into its nuclear DNA in a subpopulation of cells that synthesized DNA during the 1 hour of exposure. This is called a "pulse label" experiment. This experiment was performed for the purpose of determining whether or not the ratio of the fluorescence emission intensity of Hoechst 33342 (H42) and Syto 16 (S16) was useful in determining the amount of BrdU incorporation on a cell-by-cell basis.
[0191]
H42 and S16 stain the nuclei of live or fixed cells. The fluorescence emission of H42 disappears (decreases) when BrdU is incorporated into DNA. However, the fluorescence emission of S16 is not affected by the presence of incorporated BrdU. Thus, the ratio of H42 / S16 fluorescence intensity (calculated for each pixel using digital images) is considered to be proportional to the amount of BrdU incorporated into the DNA of each cell.
[0192]
To determine this, three culture dishes of 184B5 cells were used in these experiments. Serum-free MEBM (mammary epithelial cell basal medium, CC-3151, Clonetics) was used in the two culture dishes for the first 48 hours, and serum-free MEGM (mammary gland) was used in the third culture dish. Epithelial cell growth medium, MEGM, CC-3051, Clonetics) was used. MEGM includes bovine pituitary extract (BPE), hydrocortisone, human epidermal growth factor (EGF), and insulin, whereas MEBM does not. Thus, during the first 48 hours, culture dish 3 had all of the factors generally required for growth, while culture dishes 1 and 2 did not. MEBM only supports cells at basal metabolic levels and is not intended to support cell attachment or growth. Forty-eight hours later, MEGM was added to culture dish 2 (this two-step treatment is referred to as “MEBM + EGF”), and MEGM was also added to culture dish 3, but MEGM-hEGF (BPE, hydrocortisone, And insulin were present). The treatment of the culture dish 1 is referred to as “MEBM-EGF”.
[0193]
As shown in FIG. 4 of Stampfer et al., In 184B5 cells, DNA synthesis reached a peak about 18 hours after feeding, and the amount of DNA synthesis was highest in MEGM cells, and lower in MEBM + EGF cells. Almost no DNA synthesis was observed in the cells maintained on MEBM-EGF. These results are³H] -thymidine. In experiments performed during the development of the present invention, pulse labeling with BrdU was performed for 1 hour, and then live cells were double stained with H42 and S16. One digital image of the cells was obtained with 360 nm excitation (H42 signal) and a second image was obtained with 490 nm excitation (S16 signal). These images were processed and the ratio calculated for each pixel. The resulting ratio values were then color coded for analysis convenience. These results are based on the traditional [³[H] -Thymidine incorporation assay was consistent with that obtained. The degree of BrdU incorporation was greatest in MEGM cells (culture dish 3), followed by MEBM + EGF cells (culture dish 2), and MEBM-EGF cells (culture dish 1) showed little BrdU incorporation. .
[0194]
To further confirm these observations, cells were fixed with 70% ethanol and 50 mM glycine at pH 2.0 at -30 ° C, and indirect immunofluorescence staining was performed. After this staining, the cells were restained with H42 and S16 to determine if fixed cells labeled with anti-BrdU antibody also show a decrease in H42 fluorescence indicating BrdU disappearance. In fact, there was some evidence for this effect. MEGM cells showed the greatest amount of BrdU staining. Examination of the cells with the blue channel showed a general tendency for cells stained with BudU to have low H42 fluorescence signals. There are two ways to quantify BrdU incorporation. The first is anti-BrdU antibody detection, which is represented by the intensity of the red fluorescence. The second is to calculate, for each cell, a ratio obtained by dividing the blue nuclear fluorescence intensity (H42) by the green nuclear fluorescence intensity (S16) (that is, image ratio calculation). Since H42 is quenched by BudU incorporation but not S16, the nucleus showing red is considered to show a decrease in H42 fluorescence in the blue channel.
[0195]
Example 7
Imaging of mRNA transcription
In this embodiment, an experiment showing advantages of the VSOM experiment will be described. Such experiments would be expected to provide a means to image mRNA transcripts in tissue culture in real time, in support of the development of antisense compounds, delivery systems, and signal amplification methods. Experiments providing radiolabeled antisense compounds suitable for real-time medical imaging technology are also described.
[0196]
In these VSOM experiments, the human breast cancer cell line BT-474 is used as a model system along with mRNA encoding the bcl-2 protein. Fluorescently labeled antisense compounds in individual cells are tracked, directed to the appropriate target mRNA in the cells, and whether individual cells are affected in the expected manner. Digital imaging fluoroscopy, novel computational techniques, and the use of bioinformatics tools track fluorescent antisense compounds in one channel and monitor other cellular responses (eg, apoptosis) in another channel. Expression of bcl-2 at the mRNA level is imaged as a function of time, extracellular stress, and structure of the antisense compound. These observations are correlated with bcl-2 protein levels within each cell and the ultimate fate of each cell. Thus, these experiments provide detailed information on the biophysical properties of a particular antisense compound, useful for intelligent design of the antisense compound for imaging gene expression with PET. In fact, the in vitro techniques of the present invention allow for more compound modifications to be tested than in in vivo studies (eg, studies in nude mice transplanted with the heterologous tumor cell line BT-474).
[0197]
In some embodiments, AS-ODN compounds with the same sequence but modified at the 5 'end were used to allow both VSOM experiments and in vivo PET imaging studies. In the VSOM experiment, a fluorescent dye was used, and in the PET imaging test, one of two fluorinated radioactive components was used. In each case, a hexylamine linker was attached to the 5 'end to facilitate attachment of the fluorescent dye and radiolabel. In additional experiments, liposomes and immunoliposomes were used as delivery vehicles. In some experiments, a non-fluorescent, non-radioactive version of the fluorinated component used in the radiolabeling test was based on the biological endpoint. In this manner, the results of various substituents at the 5 'end of AS-ODN were determined. In addition, the amount of signal amplification obtained with the β-galactosidase protocol can be used with the present invention. In another embodiment, a peptide nucleic acid is used. Peptide nucleic acids (PNAs) are modified oligonucleotides in which the entire deoxyribose backbone has been replaced with a polyamide (peptide) chain (Gewirts et al., Blood 92:36 [1998]; and Temsamani and Guinot, Biotechnol. Appl. Biochem. , 26: 65-71 [1997]).
[0198]
In one embodiment, the following ASD-ODN sequence was used.

[0199]
The complete 18-mer phosphorothioate shown in SEQ ID NO: 1, having an antisense sequence for the first 6 codons of the open reading frame of bcl-2, has shown efficacy against DoHH2 lymphoma transplanted into severe immunodeficient mice (Raynaud et al., [1997]). Two 15-mer PNA-ODNs (SEQ ID NOs: 2 and 3) have been evaluated in vitro in a cell-free system (Mologni et al., [1999]). Of particular interest is that mRNA transcription is completely inhibited only when both PNA-ODNs are present simultaneously.
[0200]
Fluorescently labeled PT-G3139, PNA-1, and PNA-4 were obtained from commercial sources (eg, Genset, http://www.gensetoligos.com). These modified AS-ODNs whose backbone was completely composed of PD, PT, MP or PNA backbone bonds, as well as compounds with various substitutions and mixtures in backbone bonds were also used.
[0201]
Initial experiments were performed and recorded in the database as a series of single cell responses. A set of fluorescently labeled AS-ODNs with various modifications to the backbone binding as described above was used. A baseline VSOM run set was performed. In these runs, cells were grown on microscope coverslips and a 63 × NA1.3 oil immersion objective was used for cell observation. Cells were placed in a temperature controlled microperfusion chamber. Initial segmentation of the cytoplasmic, nuclear, and mitochondrial compartments was performed as follows.
[0202]
Computer-controlled syringe pump of fluorescent compound TMRE (tetramethyl rhodamine ethyl ester; Molecular Probes) dissolved in Dulbecco's phosphate buffered saline (D-PBS) containing calcium, magnesium, glucose, and pyruvate Into the cell chamber. Compound TMRE is a redistributable cationic dye used to determine mitochondrial membrane potential according to the Nerst equation (Farkas et al., Supra). Unlike the mitochondrial dye rhodamine 123, TMRE equilibrates rapidly and reversibly in mitochondria and, to a lesser extent, in the cytoplasm. The nucleoplasmic compartment is not stained with TMRE.
[0203]
The TMRE is then washed from the cells using a computer controlled syringe containing only buffer. TMRE can be easily washed out of the cells. Next, using another computer controlled syringe, the fluorescently labeled AS-ODN is injected into the microperfusion chamber. The concentration at which the fluorescent signal emitted by the AS-ODN becomes visible is recorded, and the perfusion of the AS-ODN is automatically stopped. This is the end of the first half of the VSOM execution.
[0204]
All digital images are added to the database for playback to confirm VSOM operation. Thus, multi-channel digital images, experimental parameters, and single-cell responses to computer-controlled compound perfusion (5-10 cells at 63 ×) become part of the database.
[0205]
In the second half of the first VSOM run, the fluorescence distribution pattern of the fluorescently labeled AS-ODN and the fluorescence distribution pattern or fluorescent AS-origin previously obtained for the same cell type using an organelle-specific fluorescent dye (Molecular Probes). A comparison is made of the cell under observation, which has been double-stained with the ODN and the organelle-specific fluorescent dye, to the current fluorescence pattern obtained in another fluorescence channel. Once the VSOM has determined the location of the AS-ODN in the cell, it adds the appropriate compound from an appropriate syringe and observes the effect on AS-ODN distribution in the cell.
[0206]
If lysosomal sequestration of the fluorescently labeled AS-ODN occurs, a compound known to selectively permeabilize the lysosome is perfused into the chamber at a controlled rate by a computer-controlled perfusion syringe. Several compounds, such as monensin proton ionophore sodium, have been shown to release fluorescent compounds trapped in acidic vesicles of human breast cancer cells (see Schindler et al., [1996]). The amount (for example) of monensin required to release the fluorescently labeled AS-ODN is recorded in a database. The time required for redistribution to adjacent compartments (eg, nuclei) and the resulting staining pattern in the nucleus is also recorded.
[0207]
VSOM perturbations to remove AS-ODN from the nucleus are performed using "clamping" techniques that are well known to those skilled in the art. These techniques allow the user to change the intracellular potassium concentration by simply changing the concentration of ions (eg, potassium) in the surrounding extracellular medium (Negulescu et al., Supra). By using this same method, intracellular calcium levels can be altered. The VSOM system progressively increases the concentration of such ions in cells because it interferes with the binding of AS-ODN to nuclear basic proteins. Again, the concentrations and other environmental conditions necessary to release the compound from the nuclear environment are recorded. If the method is not successful, online hypotonic cell lysis is performed and more stringent environmental conditions are used in subsequent experiments.
[0208]
In some embodiments, the VSOM is perturbed to release AS-ODN bound to the cell membrane by rinsing with a buffer at a non-physiological pH or with a buffer containing trypsin. Again, quantitative data on the magnitude of the perturbation required for release of the fluorescently labeled AS-ODN is recorded on a cell-by-cell basis and correlated with the chemical structure of the currently used AS-ODN.
[0209]
In a preferred embodiment, the physiological effects of AS-ODN (fluorescent and non-fluorescent) are assessed in several ways. In many cases, apoptosis-inducing extracellular stress (eg, ultraviolet light, anticancer agents, calcium ionophore, etc.) is applied. If AS-ODN successfully reduces the amount of bcl-2 protein in BT-474 cells, a decrease in bcl-2 protein and the ability of these cells to resist apoptosis are observed. In the most direct method, cells are fixed after a VSOM experiment, indirect immunofluorescence analysis using an antibody against the bcl-2 protein is performed, followed by returning to the same cell field to quantify the bcl-2 protein. (This is proportional to the fluorescence signal observed for each cell). However, it is to be understood that the invention is not limited to a particular set of steps, and that those skilled in the art will recognize that other methods are suitable for use in connection with the invention.
[0210]
Many other apoptotic markers known to those of skill in the art can also be used with the present invention. These methods include morphological and fluorescent assay systems. For example, one of the channels collected in the VSOM experiment is a transmitted light channel through which blebs of the membrane can be observed. There are also various anti-protein (eg, annexin V) antibodies and DNA dyes that can reveal nuclear fragmentation characteristic of apoptosis. In addition, dyes such as calcein-AM, which exhibit cell membrane integrity (eg, cell viability), as described in detail above, are also suitable for use in one of the fluorescent channels during the VSOM experiment. Furthermore, as already described in detail for 185b5 cells stained with LDS-751, the loss of mitochondrial membrane potential that occurs early in apoptosis is also useful as an indicator of apoptosis.
[0211]
From the foregoing, it will be apparent that the present invention provides improved methods and systems for discovering and optimizing knowledge bases on differences between cell types. In particular, the present invention provides a visual servo controlled optical microscope and analysis method. The present invention provides a means to closely monitor hundreds of individual living cells over time; quantification of dynamic physiological responses in multiple channels; real-time digital image segmentation and analysis; intelligent and repetitive, computer-applied cells. Provides stress and cell stimulation; and the ability to return to the same cell field in long-term research and observation. The invention further provides a means for optimizing culture conditions for a particular cell subpopulation.
[0212]
All publications and patents mentioned herein are incorporated herein by reference. It will be apparent to those skilled in the art that various modifications and variations can be made to the methods and systems of the present invention described herein without departing from the scope and spirit of the invention. Although the invention has been described with respect to certain preferred embodiments, it should be understood that the invention as claimed should not be unduly limited to such specific embodiments. Indeed, various modifications of the described modes for carrying out the invention which are obvious to those skilled in the art are intended to be covered by the appended claims.
[Brief description of the drawings]
FIG. 1 illustrates the relationship between the microprocessor / computer, robot, and optical system of the present invention, and other components of the system.
FIG. 2 is another diagram illustrating an alternative embodiment of the present invention.
FIG. 3 is a diagram illustrating a relationship between main objects of Blob Manager and Instrument Manager. Additional devices require plug-ins for ControlSource and BlobSource, respectively. The details of ControlSource and BlobSource are hidden at the IDL level.
4A and 4B are schematic diagrams illustrating one embodiment of a VSOM experiment.
FIG. 5 is a schematic diagram illustrating one embodiment of a multi-field (MF) VSOM experiment.
FIG. 6 provides two data models. Panel A is a complete data model for a bioinformatics system, and Panel B is a detailed breakdown for in vivo treatment.
FIG. 7 shows a process of division. Panel A provides a protocol for extracting individual nuclei from tissue and cell culture. Panel B shows how the center of gravity is converted using two adjacent nuclei.
FIG. 8 is a photograph showing original cells.
FIG. 9 shows a result of division in DAPI.
FIG. 10 illustrates the interaction of various components in one aspect of the knowledge-based VSOM of the present invention. In this figure, items in italics indicate services defined by the overall system architecture.
FIG. 11 is a diagram of detailed information on syringes, compositions in solutions in syringes, flow rates, and perfusion intervals used in some of the VSOM experiments described herein (see Example 5). is there. A pump schedule was used in both runs. After two runs (consecutively on the same day), the plots displayed on the computer screen were saved. The black window at the bottom of the figure is a plot of Run 1 and Run 2. The red line in each case represents the average of all cell responses in the entire field. As shown in these two plots recorded for MCF-7 WTC (DS, run 1) and MCF-7 (MDR, run 2), the calcein accumulation of DS cells (upper plot) (240 ± 82, N = 336) was greater than the accumulated amount of MDR cells (103 ± 80, N = 161). This is consistent with the relative differences in calcein accumulation observed for the same DS and MDR cells in multi-well plate experiments, which represent the average over many cells.
FIG. 12 is a graph showing individual cell responses of each DS cell in Run 1 of Example 4. The mean ± standard deviation of the responses of all cells is shown in black. Individual reactions are indicated by individual lines in the graph. During the VSOM experiment, the average value of the cell fluorescence of each cell was monitored and the deviation from the population average was detected. For example, arrows indicate cells that have begun to lose viability and plasma membrane integrity. Such cells cannot retain calcein and the fluorescent signal is reduced.
FIG. 13 shows the results of measuring individual cell responses for the four cells selected in Run 2 (Example 4). This is representative of the type of detailed physiological information not available with flow cytometry and multiwell plate assays, and indicates that the VSOM technology of the present invention has desirable capabilities. As shown, detailed ratio information on calcein accumulation and retention is available throughout the VSOM experiment.
FIG. 14 is a diagram of one embodiment of the present invention. Panel A is a diagram of the functional architecture of the DeepView Bioinformatics system. Panel B shows the pump log (pump
log), panel C provides an example XML recipe file, and panel D provides an ICS image header.
FIG. 15 is a flowchart of an in vivo experiment and an in vitro experiment.
FIG. 16 is a schematic diagram showing feature-based hierarchical preservation of structure and function calculated for each compartment of the cell. Each experiment consisted of time-lapse videography and the required operation of the perfusion pump (not shown). Images are stored at the lowest level. At the next level, the structure and function of the anatomy within the cell is stored as an attribute graph. At the third level, a response curve corresponding to a particular location and the expected trajectory of the compound are constructed. The fourth layer provides a mechanism for building transition probabilities along with cross-correlation tests between experiments.
FIG. 17 is a close-up schematic showing the use of an etched culture dish to establish a frame of reference for cell localization in the culture dish.

Claims (39)

A method that includes the following steps:
(A) receiving cell image data from a detection device that monitors a cell or an intracellular component of the cell;
(B) analyzing the cell image data; and (c) automatically activating a plurality of stimulators adapted to stimulate cells or intracellular components in response to the analyzed cell image data. 2. The method of claim 1, wherein (c) comprises automatically and independently operating a plurality of stimulators for stimulating cells in response to the analyzed cell image data. The method of claim 1, wherein analyzing the cell image data comprises at least one analysis selected from the group consisting of analyzing cell morphometry, analyzing cell color, and analyzing cell movement. The method of claim 1, wherein the cell image data is compared to a reference database of cell information. The method of claim 1, wherein the cell image data is compared to a remote reference database of cell information. The method of claim 1, wherein the detection device comprises a microscope, further comprising the following steps:
Monitoring cells using VSOM. The method of claim 1, wherein the stimulator is a syringe. 2. The method of claim 1, further comprising the following steps:
(D) storing the cell image data as a function of time. 2. The method of claim 1, wherein the cell is alive and the intracellular component is present in a living cell. 2. The method of claim 1, further comprising the following steps:
(D) repeating steps (a) to (c). A method for analyzing a cell array, comprising the following steps:
a) providing i) a cell array, ii) an optical system, iii) a computing means, iv) at least one test stimulus, and v) a robotic system;
b) detecting the light emitted by the cell array using an optical system to provide a first signal;
c) providing to the computing means a first signal corresponding to the light emitted by the cell array;
d) using the first signal to generate a second signal, wherein delivering the at least one test stimulus to the cell array to provide a stimulated cell array; Controlling; and e) delivering a test stimulus, wherein the optical system further detects light emitted by the stimulated cell array. The method of claim 11, wherein the test stimulus is automatically applied to the cell array according to the first signal. 12. The method of claim 11, wherein the robotic system delivers the test stimulus to the cell array via at least two microfluidic channels. The method of claim 11, wherein the optical system includes at least one microscope. 15. The method of claim 14, wherein the microscope is selected from the group consisting of a confocal microscope and a fluorescence microscope. 12. The method of claim 11, wherein the cell array comprises at least two cell types. 17. The method of claim 16, wherein the optical system scans the cell array to detect an intracellular component in a cell type included in the cell array. 18. The method of claim 17, wherein the intracellular component is selected from the group consisting of a nucleus, a nucleolus, a mitochondria, a lysosome, a phagolysosome, and a storage vesicle. 18. The method of claim 17, wherein at least one intracellular component is labeled. 19. The method of claim 18, wherein the intracellular component is labeled with the test stimulus. 18. The method of claim 17, wherein the intracellular component is labeled with at least one compound selected from the group consisting of a fluorescent dye, a radioactive compound, an enzyme, and a vital dye. The method of claim 11, wherein the optical system converts the landmark into digital data. 12. The method of claim 11, wherein the robotic system utilizes the digital data to automatically regulate the delivery of the biologically active component to the cell array. 12. The method of claim 11, further comprising analyzing the light emitted by the stimulated cell array to create a cell information profile. 26. The method of claim 24, further comprising comparing the cell information profile to a reference database of cell information. A system for analyzing a cell array, comprising:
At least one cell array, an optical system having light detection means for detecting light emitted by the cell array, a robot system, and a first program environment having means for receiving input from the optical system and output to the robot system Computer system including a second program environment having a means for providing the following. 27. The system of claim 26, wherein the robotic system further comprises at least two test stimulus sources. 27. The system of claim 26, wherein the at least two test stimulus sources include a biologically active composition. 27. The system of claim 26, wherein the at least two test stimulus sources are in fluid communication with at least one cell array. 27. The system of claim 26, wherein the cell array comprises at least two cell types. 27. The system of claim 26, further comprising means for remotely operating a computer system. 27. The system of claim 26, wherein the optical system includes a microscope. Computer readable media, including:
(A) a code for receiving cell image data from a detector that monitors cells or intracellular components;
(B) code for analyzing cell image data; and (c) for automatically adjusting a plurality of stimulators adapted to stimulate cells or intracellular components in response to the analyzed cell image data. Code. 34. The computer readable medium of claim 33, further comprising:
(D) Code for controlling a visual servo controlled light microscope (VSOM) to monitor cells. 34. The computer readable medium of claim 33, further comprising at least one of the following:
(D) a code for analyzing cell size, a code for analyzing cell color, and a code for analyzing cell movement. 34. The computer readable medium of claim 33, further comprising:
(D) Code for automatically and independently operating a plurality of stimulators for stimulating cells according to the analyzed cell image data. 34. The computer readable medium according to claim 33, wherein the stimulator is a syringe. 34. The computer-readable medium of claim 33, further comprising code for storing cell image data as a function of time. 34. The computer readable medium of claim 33, wherein the cells are alive and the intracellular components are present in living cells.

Images